'AVG CPU Busy',
'Average CPU busy across all hosts (OSD, RGW, MON etc) within the cluster',
'current',
- 'avg(\n 1 - (\n avg by(instance) \n (irate(node_cpu_seconds_total{mode=\'idle\',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]) or\n irate(node_cpu{mode=\'idle\',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]))\n )\n )',
+ |||
+ avg(1 - (
+ avg by(instance) (
+ irate(node_cpu_seconds_total{mode=\'idle\',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]) or
+ irate(node_cpu{mode=\'idle\',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m])
+ )
+ ))
+ |||,
true,
4,
0,
'AVG RAM Utilization',
'Average Memory Usage across all hosts in the cluster (excludes buffer/cache usage)',
'current',
- 'avg (((node_memory_MemTotal{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or node_memory_MemTotal_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"})- (\n (node_memory_MemFree{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or node_memory_MemFree_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"}) + \n (node_memory_Cached{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or node_memory_Cached_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"}) + \n (node_memory_Buffers{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or node_memory_Buffers_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"}) +\n (node_memory_Slab{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or node_memory_Slab_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"})\n )) /\n (node_memory_MemTotal{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or node_memory_MemTotal_bytes{instance=~"($osd_hosts|$rgw_hosts|$mon_hosts|$mds_hosts).*"} ))',
+ |||
+ avg ((
+ (
+ node_memory_MemTotal{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or
+ node_memory_MemTotal_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"}
+ ) - ((
+ node_memory_MemFree{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or
+ node_memory_MemFree_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"}) +
+ (
+ node_memory_Cached{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or
+ node_memory_Cached_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"}
+ ) + (
+ node_memory_Buffers{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or
+ node_memory_Buffers_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"}
+ ) + (
+ node_memory_Slab{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or
+ node_memory_Slab_bytes{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"}
+ )
+ )
+ )
+ (
+ node_memory_MemTotal{instance=~"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*"} or
+ node_memory_MemTotal_bytes{instance=~"($osd_hosts|$rgw_hosts|$mon_hosts|$mds_hosts).*"}
+ ))
+ |||,
true,
8,
0,
'Physical IOPS',
'IOPS Load at the device as reported by the OS on all OSD hosts',
'current',
- 'sum ((irate(node_disk_reads_completed{instance=~"($osd_hosts).*"}[5m]) or irate(node_disk_reads_completed_total{instance=~"($osd_hosts).*"}[5m]) ) + \n(irate(node_disk_writes_completed{instance=~"($osd_hosts).*"}[5m]) or irate(node_disk_writes_completed_total{instance=~"($osd_hosts).*"}[5m])))',
+ |||
+ sum ((
+ irate(node_disk_reads_completed{instance=~"($osd_hosts).*"}[5m]) or
+ irate(node_disk_reads_completed_total{instance=~"($osd_hosts).*"}[5m])
+ ) + (
+ irate(node_disk_writes_completed{instance=~"($osd_hosts).*"}[5m]) or
+ irate(node_disk_writes_completed_total{instance=~"($osd_hosts).*"}[5m])
+ ))
+ |||,
true,
12,
0,
'AVG Disk Utilization',
'Average Disk utilization for all OSD data devices (i.e. excludes journal/WAL)',
'current',
- 'avg (\n label_replace((irate(node_disk_io_time_ms[5m]) / 10 ) or\n (irate(node_disk_io_time_seconds_total[5m]) * 100), "instance", "$1", "instance", "([^.:]*).*"\n ) *\n on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human{instance=~"($osd_hosts).*"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^.:]*).*")\n)',
+ |||
+ avg (
+ label_replace(
+ (irate(node_disk_io_time_ms[5m]) / 10 ) or
+ (irate(node_disk_io_time_seconds_total[5m]) * 100),
+ "instance", "$1", "instance", "([^.:]*).*"
+ ) * on(instance, device) group_left(ceph_daemon) label_replace(
+ label_replace(
+ ceph_disk_occupation_human{instance=~"($osd_hosts).*"},
+ "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^.:]*).*"
+ )
+ )
+ |||,
true,
16,
0,
'CPU Busy - Top 10 Hosts',
'Show the top 10 busiest hosts by cpu',
'percent',
- 'topk(10,100 * ( 1 - (\n avg by(instance) \n (irate(node_cpu_seconds_total{mode=\'idle\',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]) or\n irate(node_cpu{mode=\'idle\',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]))\n )\n )\n)',
+ |||
+ topk(10,
+ 100 * (
+ 1 - (
+ avg by(instance) (
+ irate(node_cpu_seconds_total{mode=\'idle\',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]) or
+ irate(node_cpu{mode=\'idle\',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m])
+ )
+ )
+ )
+ )
+ |||,
'{{instance}}',
0,
5,
system: '#890F02',
user: '#3F6833',
wait: '#C15C17',
- }, 'CPU Utilization', "Shows the CPU breakdown. When multiple servers are selected, only the first host's cpu data is shown", 'null', 'percent', '% Utilization', 'sum by (mode) (\n irate(node_cpu{instance=~"($ceph_hosts)([\\\\.:].*)?", mode=~"(irq|nice|softirq|steal|system|user|iowait)"}[1m]) or\n irate(node_cpu_seconds_total{instance=~"($ceph_hosts)([\\\\.:].*)?", mode=~"(irq|nice|softirq|steal|system|user|iowait)"}[1m])\n) / scalar(\n sum(irate(node_cpu{instance=~"($ceph_hosts)([\\\\.:].*)?"}[1m]) or\n irate(node_cpu_seconds_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[1m]))\n) * 100', '{{mode}}', 3, 1, 6, 10
+ },
+ 'CPU Utilization',
+ "Shows the CPU breakdown. When multiple servers are selected, only the first host's cpu data is shown",
+ 'null',
+ 'percent',
+ '% Utilization',
+ |||
+ sum by (mode) (
+ irate(node_cpu{instance=~"($ceph_hosts)([\\\\.:].*)?", mode=~"(irq|nice|softirq|steal|system|user|iowait)"}[1m]) or
+ irate(node_cpu_seconds_total{instance=~"($ceph_hosts)([\\\\.:].*)?", mode=~"(irq|nice|softirq|steal|system|user|iowait)"}[1m])
+ ) / (
+ scalar(
+ sum(irate(node_cpu{instance=~"($ceph_hosts)([\\\\.:].*)?"}[1m]) or
+ irate(node_cpu_seconds_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[1m]))
+ ) * 100
+ )
+ |||,
+ '{{mode}}',
+ 3,
+ 1,
+ 6,
+ 10
),
HostDetailsGraphPanel(
{
'null',
'bytes',
'RAM used',
- 'node_memory_MemFree{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_MemFree_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"} ',
+ |||
+ node_memory_MemFree{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_MemFree_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ |||,
'Free',
9,
1,
)
.addTargets(
[
- u.addTargetSchema('node_memory_MemTotal{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_MemTotal_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"} ', 'total'),
- u.addTargetSchema('(node_memory_Cached{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_Cached_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}) + \n(node_memory_Buffers{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_Buffers_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}) +\n(node_memory_Slab{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_Slab_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}) \n', 'buffers/cache'),
- u.addTargetSchema('(node_memory_MemTotal{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_MemTotal_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"})- (\n (node_memory_MemFree{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_MemFree_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}) + \n (node_memory_Cached{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_Cached_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}) + \n (node_memory_Buffers{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_Buffers_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}) +\n (node_memory_Slab{instance=~"$ceph_hosts([\\\\.:].*)?"} or node_memory_Slab_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"})\n )\n \n', 'used'),
+ u.addTargetSchema(
+ |||
+ node_memory_MemTotal{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_MemTotal_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ |||,
+ 'total'
+ ),
+ u.addTargetSchema(
+ |||
+ (
+ node_memory_Cached{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_Cached_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ ) + (
+ node_memory_Buffers{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_Buffers_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ ) + (
+ node_memory_Slab{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_Slab_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ )
+ |||,
+ 'buffers/cache'
+ ),
+ u.addTargetSchema(
+ |||
+ (
+ node_memory_MemTotal{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_MemTotal_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ ) - (
+ (
+ node_memory_MemFree{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_MemFree_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ ) + (
+ node_memory_Cached{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_Cached_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ ) + (
+ node_memory_Buffers{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_Buffers_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ ) +
+ (
+ node_memory_Slab{instance=~"$ceph_hosts([\\\\.:].*)?"} or
+ node_memory_Slab_bytes{instance=~"$ceph_hosts([\\\\.:].*)?"}
+ )
+ )
+ |||,
+ 'used'
+ ),
]
)
.addSeriesOverride(
'null',
'decbytes',
'Send (-) / Receive (+)',
- 'sum by (device) (\n irate(node_network_receive_bytes{instance=~"($ceph_hosts)([\\\\.:].*)?",device!="lo"}[1m]) or \n irate(node_network_receive_bytes_total{instance=~"($ceph_hosts)([\\\\.:].*)?",device!="lo"}[1m])\n)',
+ |||
+ sum by (device) (
+ irate(
+ node_network_receive_bytes{instance=~"($ceph_hosts)([\\\\.:].*)?",device!="lo"}[1m]) or
+ irate(node_network_receive_bytes_total{instance=~"($ceph_hosts)([\\\\.:].*)?",device!="lo"}[1m]
+ )
+ )
+ |||,
'{{device}}.rx',
15,
1,
)
.addTargets(
[
- u.addTargetSchema('sum by (device) (\n irate(node_network_transmit_bytes{instance=~"($ceph_hosts)([\\\\.:].*)?",device!="lo"}[1m]) or\n irate(node_network_transmit_bytes_total{instance=~"($ceph_hosts)([\\\\.:].*)?",device!="lo"}[1m])\n)', '{{device}}.tx'),
+ u.addTargetSchema(
+ |||
+ sum by (device) (
+ irate(node_network_transmit_bytes{instance=~"($ceph_hosts)([\\\\.:].*)?",device!="lo"}[1m]) or
+ irate(node_network_transmit_bytes_total{instance=~"($ceph_hosts)([\\\\.:].*)?",device!="lo"}[1m])
+ )
+ |||,
+ '{{device}}.tx'
+ ),
]
)
.addSeriesOverride(
'null',
'pps',
'Send (-) / Receive (+)',
- 'irate(node_network_receive_drop{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m]) or irate(node_network_receive_drop_total{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m])',
+ |||
+ irate(node_network_receive_drop{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m]) or
+ irate(node_network_receive_drop_total{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m])
+ |||,
'{{device}}.rx',
21,
1,
.addTargets(
[
u.addTargetSchema(
- 'irate(node_network_transmit_drop{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m]) or irate(node_network_transmit_drop_total{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m])', '{{device}}.tx'
+ |||
+ irate(node_network_transmit_drop{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m]) or
+ irate(node_network_transmit_drop_total{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m])
+ |||,
+ '{{device}}.tx'
),
]
)
'Raw Capacity',
'Each OSD consists of a Journal/WAL partition and a data partition. The RAW Capacity shown is the sum of the data partitions across all OSDs on the selected OSD hosts.',
'current',
- 'sum(ceph_osd_stat_bytes and on (ceph_daemon) ceph_disk_occupation{instance=~"($ceph_hosts)([\\\\.:].*)?"})',
+ |||
+ sum(
+ ceph_osd_stat_bytes and
+ on (ceph_daemon) ceph_disk_occupation{instance=~"($ceph_hosts)([\\\\.:].*)?"}
+ )
+ |||,
0,
6,
3,
'null',
'pps',
'Send (-) / Receive (+)',
- 'irate(node_network_receive_errs{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m]) or irate(node_network_receive_errs_total{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m])',
+ |||
+ irate(node_network_receive_errs{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m]) or
+ irate(node_network_receive_errs_total{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m])
+ |||,
'{{device}}.rx',
21,
6,
)
.addTargets(
[u.addTargetSchema(
- 'irate(node_network_transmit_errs{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m]) or irate(node_network_transmit_errs_total{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m])', '{{device}}.tx'
+ |||
+ irate(node_network_transmit_errs{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m]) or
+ irate(node_network_transmit_errs_total{instance=~"$ceph_hosts([\\\\.:].*)?"}[1m])
+ |||,
+ '{{device}}.tx'
)]
)
.addSeriesOverride(
'connected',
'ops',
'Read (-) / Write (+)',
- 'label_replace(\n (\n irate(node_disk_writes_completed{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) or\n irate(node_disk_writes_completed_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])\n ),\n "instance",\n "$1",\n "instance",\n "([^:.]*).*"\n)\n* on(instance, device) group_left(ceph_daemon)\n label_replace(\n label_replace(\n ceph_disk_occupation_human,\n "device",\n "$1",\n "device",\n "/dev/(.*)"\n ),\n "instance",\n "$1",\n "instance",\n "([^:.]*).*"\n )',
+ |||
+ label_replace(
+ (
+ irate(node_disk_writes_completed{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) or
+ irate(node_disk_writes_completed_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ ) * on(instance, device) group_left(ceph_daemon) label_replace(
+ label_replace(
+ ceph_disk_occupation_human, "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
'{{device}}({{ceph_daemon}}) writes',
0,
12,
.addTargets(
[
u.addTargetSchema(
- 'label_replace(\n (irate(node_disk_reads_completed{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) or irate(node_disk_reads_completed_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])),\n "instance",\n "$1",\n "instance",\n "([^:.]*).*"\n)\n* on(instance, device) group_left(ceph_daemon)\n label_replace(\n label_replace(\n ceph_disk_occupation_human,\n "device",\n "$1",\n "device",\n "/dev/(.*)"\n ),\n "instance",\n "$1",\n "instance",\n "([^:.]*).*"\n )',
+ |||
+ label_replace(
+ (
+ irate(node_disk_reads_completed{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) or
+ irate(node_disk_reads_completed_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ ) * on(instance, device) group_left(ceph_daemon) label_replace(
+ label_replace(
+ ceph_disk_occupation_human,"device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
'{{device}}({{ceph_daemon}}) reads'
),
]
'connected',
'Bps',
'Read (-) / Write (+)',
- 'label_replace((irate(node_disk_bytes_written{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) or irate(node_disk_written_bytes_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])), "instance", "$1", "instance", "([^:.]*).*") * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*")',
+ |||
+ label_replace(
+ (
+ irate(node_disk_bytes_written{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) or
+ irate(node_disk_written_bytes_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])
+ ), "instance", "$1", "instance", "([^:.]*).*") * on(instance, device)
+ group_left(ceph_daemon) label_replace(
+ label_replace(ceph_disk_occupation_human, "device", "$1", "device", "/dev/(.*)"),
+ "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
'{{device}}({{ceph_daemon}}) write',
12,
12,
)
.addTargets(
[u.addTargetSchema(
- 'label_replace((irate(node_disk_bytes_read{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) or irate(node_disk_read_bytes_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])), "instance", "$1", "instance", "([^:.]*).*") * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*")',
+ |||
+ label_replace(
+ (
+ irate(node_disk_bytes_read{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) or
+ irate(node_disk_read_bytes_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])
+ ),
+ "instance", "$1", "instance", "([^:.]*).*") * on(instance, device)
+ group_left(ceph_daemon) label_replace(
+ label_replace(ceph_disk_occupation_human, "device", "$1", "device", "/dev/(.*)"),
+ "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
'{{device}}({{ceph_daemon}}) read'
)]
)
'null as zero',
's',
'',
- 'max by(instance,device) (label_replace((irate(node_disk_write_time_seconds_total{ instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) ) / clamp_min(irate(node_disk_writes_completed_total{ instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]), 0.001) or (irate(node_disk_read_time_seconds_total{ instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) ) / clamp_min(irate(node_disk_reads_completed_total{ instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]), 0.001), "instance", "$1", "instance", "([^:.]*).*")) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human{instance=~"($ceph_hosts)([\\\\.:].*)?"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*")',
+ |||
+ max by(instance, device) (label_replace(
+ (irate(node_disk_write_time_seconds_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])) /
+ clamp_min(irate(node_disk_writes_completed_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]), 0.001) or
+ (irate(node_disk_read_time_seconds_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m])) /
+ clamp_min(irate(node_disk_reads_completed_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]), 0.001),
+ "instance", "$1", "instance", "([^:.]*).*"
+ )) * on(instance, device) group_left(ceph_daemon) label_replace(
+ label_replace(
+ ceph_disk_occupation_human{instance=~"($ceph_hosts)([\\\\.:].*)?"},
+ "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
'{{device}}({{ceph_daemon}})',
0,
21,
'connected',
'percent',
'%Util',
- 'label_replace(((irate(node_disk_io_time_ms{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) / 10 ) or irate(node_disk_io_time_seconds_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) * 100), "instance", "$1", "instance", "([^:.]*).*") * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human{instance=~"($ceph_hosts)([\\\\.:].*)?"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*")',
+ |||
+ label_replace(
+ (
+ (irate(node_disk_io_time_ms{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) / 10) or
+ irate(node_disk_io_time_seconds_total{instance=~"($ceph_hosts)([\\\\.:].*)?"}[5m]) * 100
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ ) * on(instance, device) group_left(ceph_daemon) label_replace(
+ label_replace(ceph_disk_occupation_human{instance=~"($ceph_hosts)([\\\\.:].*)?"},
+ "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
'{{device}}({{ceph_daemon}})',
12,
21,
.addTargets(
[
u.addTargetSchema(
- 'max (irate(ceph_osd_op_r_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000)',
+ |||
+ max(
+ irate(ceph_osd_op_r_latency_sum[1m]) /
+ on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000
+ )
+ |||,
'MAX read'
),
u.addTargetSchema(
- 'quantile(0.95,\n (irate(ceph_osd_op_r_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000)\n)', '@95%ile'
+ |||
+ quantile(0.95,
+ (
+ irate(ceph_osd_op_r_latency_sum[1m]) /
+ on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m])
+ * 1000
+ )
+ )
+ |||,
+ '@95%ile'
),
],
),
)
.addTarget(
u.addTargetSchema(
- 'topk(10,\n (sort(\n (irate(ceph_osd_op_r_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000)\n ))\n)\n\n', '', 'table', 1, true
+ |||
+ topk(10,
+ (sort(
+ (
+ irate(ceph_osd_op_r_latency_sum[1m]) /
+ on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) *
+ 1000
+ )
+ ))
+ )
+ |||,
+ '',
+ 'table',
+ 1,
+ true
)
) + { gridPos: { x: 8, y: 0, w: 4, h: 8 } },
OsdOverviewGraphPanel(
'ms',
null,
'0',
- 'avg (irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000)',
+ |||
+ avg(
+ irate(ceph_osd_op_w_latency_sum[1m]) /
+ on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m])
+ * 1000
+ )
+ |||,
'AVG write',
12,
0,
.addTargets(
[
u.addTargetSchema(
- 'max (irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000)',
- 'MAX write'
+ |||
+ max(
+ irate(ceph_osd_op_w_latency_sum[1m]) /
+ on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) *
+ 1000
+ )
+ |||, 'MAX write'
),
u.addTargetSchema(
- 'quantile(0.95,\n (irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000)\n)', '@95%ile write'
+ |||
+ quantile(0.95, (
+ irate(ceph_osd_op_w_latency_sum[1m]) /
+ on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) *
+ 1000
+ ))
+ |||, '@95%ile write'
),
],
),
)
.addTarget(
u.addTargetSchema(
- 'topk(10,\n (sort(\n (irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000)\n ))\n)\n\n',
+ |||
+ topk(10,
+ (sort(
+ (irate(ceph_osd_op_w_latency_sum[1m]) /
+ on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) *
+ 1000)
+ ))
+ )
+ |||,
'',
'table',
1,
min='0',
nullPointMode='null')
.addTarget(u.addTargetSchema(
- 'ceph_osd_numpg\n', 'PGs per OSD', 'time_series', 1, true
+ 'ceph_osd_numpg', 'PGs per OSD', 'time_series', 1, true
)) + { gridPos: { x: 12, y: 8, w: 8, h: 8 } },
OsdOverviewSingleStatPanel(
['#d44a3a', '#299c46'],
true,
false,
'.75',
- 'sum(ceph_bluestore_onode_hits)/(sum(ceph_bluestore_onode_hits) + sum(ceph_bluestore_onode_misses))',
+ |||
+ sum(ceph_bluestore_onode_hits) / (
+ sum(ceph_bluestore_onode_hits) +
+ sum(ceph_bluestore_onode_misses)
+ )
+ |||,
20,
8,
4,
'',
's',
'Read (-) / Write (+)',
- 'irate(ceph_osd_op_r_latency_sum{ceph_daemon=~"$osd"}[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m])',
- 'irate(ceph_osd_op_w_latency_sum{ceph_daemon=~"$osd"}[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m])',
+ |||
+ irate(ceph_osd_op_r_latency_sum{ceph_daemon=~"$osd"}[1m]) /
+ on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m])
+ |||,
+ |||
+ irate(ceph_osd_op_w_latency_sum{ceph_daemon=~"$osd"}[1m]) /
+ on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m])
+ |||,
'read',
'write',
0,
'',
's',
'Read (-) / Write (+)',
- '(label_replace(irate(node_disk_read_time_seconds_total[1m]) / irate(node_disk_reads_completed_total[1m]), "instance", "$1", "instance", "([^:.]*).*") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~"$osd"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*"))',
- '(label_replace(irate(node_disk_write_time_seconds_total[1m]) / irate(node_disk_writes_completed_total[1m]), "instance", "$1", "instance", "([^:.]*).*") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~"$osd"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*"))',
+ |||
+ (
+ label_replace(
+ irate(node_disk_read_time_seconds_total[1m]) / irate(node_disk_reads_completed_total[1m]),
+ "instance", "$1", "instance", "([^:.]*).*"
+ ) and on (instance, device) label_replace(
+ label_replace(
+ ceph_disk_occupation_human{ceph_daemon=~"$osd"},
+ "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ )
+ |||,
+ |||
+ (
+ label_replace(
+ irate(node_disk_write_time_seconds_total[1m]) / irate(node_disk_writes_completed_total[1m]),
+ "instance", "$1", "instance", "([^:.]*).*") and on (instance, device)
+ label_replace(
+ label_replace(
+ ceph_disk_occupation_human{ceph_daemon=~"$osd"}, "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ )
+ |||,
'{{instance}}/{{device}} Reads',
'{{instance}}/{{device}} Writes',
0,
'',
'short',
'Read (-) / Write (+)',
- 'label_replace(irate(node_disk_writes_completed_total[1m]), "instance", "$1", "instance", "([^:.]*).*") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~"$osd"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*")',
- 'label_replace(irate(node_disk_reads_completed_total[1m]), "instance", "$1", "instance", "([^:.]*).*") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~"$osd"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*")',
+ |||
+ label_replace(
+ irate(node_disk_writes_completed_total[1m]),
+ "instance", "$1", "instance", "([^:.]*).*"
+ ) and on (instance, device) label_replace(
+ label_replace(
+ ceph_disk_occupation_human{ceph_daemon=~"$osd"},
+ "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
+ |||
+ label_replace(
+ irate(node_disk_reads_completed_total[1m]),
+ "instance", "$1", "instance", "([^:.]*).*"
+ ) and on (instance, device) label_replace(
+ label_replace(
+ ceph_disk_occupation_human{ceph_daemon=~"$osd"},
+ "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
'{{device}} on {{instance}} Writes',
'{{device}} on {{instance}} Reads',
6,
'',
'Bps',
'Read (-) / Write (+)',
- 'label_replace(irate(node_disk_read_bytes_total[1m]), "instance", "$1", "instance", "([^:.]*).*") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~"$osd"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*")',
- 'label_replace(irate(node_disk_written_bytes_total[1m]), "instance", "$1", "instance", "([^:.]*).*") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~"$osd"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*")',
+ |||
+ label_replace(
+ irate(node_disk_read_bytes_total[1m]), "instance", "$1", "instance", "([^:.]*).*"
+ ) and on (instance, device) label_replace(
+ label_replace(
+ ceph_disk_occupation_human{ceph_daemon=~"$osd"},
+ "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
+ |||
+ label_replace(
+ irate(node_disk_written_bytes_total[1m]), "instance", "$1", "instance", "([^:.]*).*"
+ ) and on (instance, device) label_replace(
+ label_replace(
+ ceph_disk_occupation_human{ceph_daemon=~"$osd"},
+ "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
'{{instance}} {{device}} Reads',
'{{instance}} {{device}} Writes',
12,
'$datasource'
)
.addTarget(u.addTargetSchema(
- 'label_replace(irate(node_disk_io_time_seconds_total[1m]), "instance", "$1", "instance", "([^:.]*).*") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~"$osd"}, "device", "$1", "device", "/dev/(.*)"), "instance", "$1", "instance", "([^:.]*).*")',
+ |||
+ label_replace(
+ irate(node_disk_io_time_seconds_total[1m]),
+ "instance", "$1", "instance", "([^:.]*).*"
+ ) and on (instance, device) label_replace(
+ label_replace(
+ ceph_disk_occupation_human{ceph_daemon=~"$osd"}, "device", "$1", "device", "/dev/(.*)"
+ ), "instance", "$1", "instance", "([^:.]*).*"
+ )
+ |||,
'{{device}} on {{instance}}'
)) + { gridPos: { x: 18, y: 11, w: 6, h: 9 } },
]),
22,
[],
'',
- { refresh_intervals: ['5s', '10s', '15s', '30s', '1m', '5m', '15m', '30m', '1h', '2h', '1d'], time_options: ['5m', '15m', '1h', '6h', '12h', '24h', '2d', '7d', '30d'] }
+ {
+ refresh_intervals: ['5s', '10s', '15s', '30s', '1m', '5m', '15m', '30m', '1h', '2h', '1d'],
+ time_options: ['5m', '15m', '1h', '6h', '12h', '24h', '2d', '7d', '30d'],
+ }
)
.addAnnotation(
u.addAnnotationSchema(
PoolOverviewSingleStatPanel(
'percent',
'Compression Eligibility',
- 'Indicates how suitable the data is within the pools that are/have been enabled for compression - averaged across all pools holding compressed data\n',
+ 'Indicates how suitable the data is within the pools that are/have been enabled for compression - averaged across all pools holding compressed data',
'current',
- '(sum(ceph_pool_compress_under_bytes > 0) / sum(ceph_pool_stored_raw and ceph_pool_compress_under_bytes > 0)) * 100',
+ |||
+ (
+ sum(ceph_pool_compress_under_bytes > 0) /
+ sum(ceph_pool_stored_raw and ceph_pool_compress_under_bytes > 0)
+ ) * 100
+ |||,
null,
'table',
18,
.addTargets(
[
u.addTargetSchema(
- '(ceph_pool_compress_under_bytes / ceph_pool_compress_bytes_used > 0) and on(pool_id) (((ceph_pool_compress_under_bytes > 0) / ceph_pool_stored_raw) * 100 > 0.5)',
+ |||
+ (ceph_pool_compress_under_bytes / ceph_pool_compress_bytes_used > 0) and on(pool_id) (
+ ((ceph_pool_compress_under_bytes > 0) / ceph_pool_stored_raw) * 100 > 0.5
+ )
+ |||,
'A',
'table',
1,
'This chart shows the sum of read and write IOPS from all clients by pool',
'short',
'IOPS',
- 'topk($topk,round((rate(ceph_pool_rd[30s]) + rate(ceph_pool_wr[30s])),1) * on(pool_id) group_left(instance,name) ceph_pool_metadata) ',
+ |||
+ topk($topk,
+ round(
+ (rate(ceph_pool_rd[30s]) + rate(ceph_pool_wr[30s])),
+ 1
+ ) * on(pool_id) group_left(instance,name) ceph_pool_metadata)
+ |||,
'{{name}} ',
0,
9,
)
.addTarget(
u.addTargetSchema(
- 'topk($topk,rate(ceph_pool_wr[30s]) + on(pool_id) group_left(instance,name) ceph_pool_metadata) ',
+ |||
+ topk($topk,
+ rate(ceph_pool_wr[30s]) +
+ on(pool_id) group_left(instance,name) ceph_pool_metadata
+ )
+ |||,
'{{name}} - write'
)
),
'The chart shows the sum of read and write bytes from all clients, by pool',
'Bps',
'Throughput',
- 'topk($topk,(rate(ceph_pool_rd_bytes[30s]) + rate(ceph_pool_wr_bytes[30s])) * on(pool_id) group_left(instance,name) ceph_pool_metadata)',
+ |||
+ topk($topk,
+ (rate(ceph_pool_rd_bytes[30s]) + rate(ceph_pool_wr_bytes[30s])) *
+ on(pool_id) group_left(instance, name) ceph_pool_metadata
+ )
+ |||,
'{{name}}',
12,
9,
true,
true,
'.7,.8',
- '(ceph_pool_stored / (ceph_pool_stored + ceph_pool_max_avail)) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}',
+ |||
+ (ceph_pool_stored / (ceph_pool_stored + ceph_pool_max_avail)) *
+ on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}
+ |||,
'time_series',
0,
0,
false,
'',
'current',
- '(ceph_pool_max_avail / deriv(ceph_pool_stored[6h])) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"} > 0',
+ |||
+ (ceph_pool_max_avail / deriv(ceph_pool_stored[6h])) *
+ on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"} > 0
+ |||,
'time_series',
7,
0,
'',
'ops',
'Objects out(-) / in(+) ',
- 'deriv(ceph_pool_objects[1m]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}',
+ |||
+ deriv(ceph_pool_objects[1m]) *
+ on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}
+ |||,
'Objects per second',
12,
0,
{
read_op_per_sec: '#3F6833',
write_op_per_sec: '#E5AC0E',
- }, '$pool_name Client IOPS', '', 'iops', 'Read (-) / Write (+)', 'irate(ceph_pool_rd[1m]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}', 'reads', 0, 7, 12, 7
+ },
+ '$pool_name Client IOPS',
+ '',
+ 'iops',
+ 'Read (-) / Write (+)',
+ |||
+ irate(ceph_pool_rd[1m]) *
+ on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}
+ |||,
+ 'reads',
+ 0,
+ 7,
+ 12,
+ 7
)
.addSeriesOverride({ alias: 'reads', transform: 'negative-Y' })
.addTarget(
u.addTargetSchema(
- 'irate(ceph_pool_wr[1m]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}', 'writes'
+ |||
+ irate(ceph_pool_wr[1m]) *
+ on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}
+ |||,
+ 'writes'
)
),
PoolDetailGraphPanel(
'',
'Bps',
'Read (-) / Write (+)',
- 'irate(ceph_pool_rd_bytes[1m]) + on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}',
+ |||
+ irate(ceph_pool_rd_bytes[1m]) +
+ on(pool_id) group_left(instance, name) ceph_pool_metadata{name=~"$pool_name"}
+ |||,
'reads',
12,
7,
.addSeriesOverride({ alias: 'reads', transform: 'negative-Y' })
.addTarget(
u.addTargetSchema(
- 'irate(ceph_pool_wr_bytes[1m]) + on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}',
+ |||
+ irate(ceph_pool_wr_bytes[1m]) +
+ on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}
+ |||,
'writes'
)
),
'',
'short',
'Objects',
- 'ceph_pool_objects * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}',
+ |||
+ ceph_pool_objects *
+ on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~"$pool_name"}
+ |||,
'Number of Objects',
0,
14,
RbdDetailsPanel(
'Average Latency',
'ns',
- 'irate(ceph_rbd_write_latency_sum{pool="$Pool", image="$Image"}[30s]) / irate(ceph_rbd_write_latency_count{pool="$Pool", image="$Image"}[30s])',
- 'irate(ceph_rbd_read_latency_sum{pool="$Pool", image="$Image"}[30s]) / irate(ceph_rbd_read_latency_count{pool="$Pool", image="$Image"}[30s])',
+ |||
+ irate(ceph_rbd_write_latency_sum{pool="$Pool", image="$Image"}[30s]) /
+ irate(ceph_rbd_write_latency_count{pool="$Pool", image="$Image"}[30s])
+ |||,
+ |||
+ irate(ceph_rbd_read_latency_sum{pool="$Pool", image="$Image"}[30s]) /
+ irate(ceph_rbd_read_latency_count{pool="$Pool", image="$Image"}[30s])
+ |||,
16,
0,
8,
RbdOverviewPanel(
'Average Latency',
'ns',
- 'round(sum(irate(ceph_rbd_write_latency_sum[30s])) / sum(irate(ceph_rbd_write_latency_count[30s])))',
- 'round(sum(irate(ceph_rbd_read_latency_sum[30s])) / sum(irate(ceph_rbd_read_latency_count[30s])))',
+ |||
+ round(
+ sum(irate(ceph_rbd_write_latency_sum[30s])) /
+ sum(irate(ceph_rbd_write_latency_count[30s]))
+ )
+ |||,
+ |||
+ round(
+ sum(irate(ceph_rbd_read_latency_sum[30s])) /
+ sum(irate(ceph_rbd_read_latency_count[30s]))
+ )
+ |||,
'Write',
'Read',
16,
)
.addTarget(
u.addTargetSchema(
- 'topk(10, (sort((irate(ceph_rbd_write_ops[30s]) + on (image, pool, namespace) irate(ceph_rbd_read_ops[30s])))))',
+ |||
+ topk(10,
+ (
+ sort((
+ irate(ceph_rbd_write_ops[30s]) +
+ on (image, pool, namespace) irate(ceph_rbd_read_ops[30s])
+ ))
+ )
+ )
+ |||,
'',
'table',
1,
)
.addTarget(
u.addTargetSchema(
- 'topk(10, sort(sum(irate(ceph_rbd_read_bytes[30s]) + irate(ceph_rbd_write_bytes[30s])) by (pool, image, namespace)))',
+ |||
+ topk(10,
+ sort(
+ sum(
+ irate(ceph_rbd_read_bytes[30s]) + irate(ceph_rbd_write_bytes[30s])
+ ) by (pool, image, namespace)
+ )
+ )
+ |||,
'',
'table',
1,
)
.addTarget(
u.addTargetSchema(
- 'topk(10,\n sum(\n irate(ceph_rbd_write_latency_sum[30s]) / clamp_min(irate(ceph_rbd_write_latency_count[30s]), 1) +\n irate(ceph_rbd_read_latency_sum[30s]) / clamp_min(irate(ceph_rbd_read_latency_count[30s]), 1)\n ) by (pool, image, namespace)\n)',
+ |||
+ topk(10,
+ sum(
+ irate(ceph_rbd_write_latency_sum[30s]) / clamp_min(irate(ceph_rbd_write_latency_count[30s]), 1) +
+ irate(ceph_rbd_read_latency_sum[30s]) / clamp_min(irate(ceph_rbd_read_latency_count[30s]), 1)
+ ) by (pool, image, namespace)
+ )
+ |||,
'',
'table',
1,
'',
's',
'short',
- 'rate(ceph_rgw_get_initial_lat_sum[30s]) / rate(ceph_rgw_get_initial_lat_count[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata',
+ |||
+ rate(ceph_rgw_get_initial_lat_sum[30s]) /
+ rate(ceph_rgw_get_initial_lat_count[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata
+ |||,
'GET AVG',
0,
1,
).addTargets(
[
u.addTargetSchema(
- 'rate(ceph_rgw_put_initial_lat_sum[30s]) / rate(ceph_rgw_put_initial_lat_count[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata',
+ |||
+ rate(ceph_rgw_put_initial_lat_sum[30s]) /
+ rate(ceph_rgw_put_initial_lat_count[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata
+ |||,
'PUT AVG'
),
]
'',
'none',
'short',
- 'sum by (rgw_host) (label_replace(rate(ceph_rgw_req[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata, "rgw_host", "$1", "ceph_daemon", "rgw.(.*)"))',
+ |||
+ sum by (rgw_host) (
+ label_replace(
+ rate(ceph_rgw_req[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,
+ "rgw_host", "$1", "ceph_daemon", "rgw.(.*)"
+ )
+ )
+ |||,
'{{rgw_host}}',
8,
1,
'Latencies are shown stacked, without a yaxis to provide a visual indication of GET latency imbalance across RGW hosts',
's',
'short',
- 'label_replace(\n rate(ceph_rgw_get_initial_lat_sum[30s]) /\n rate(ceph_rgw_get_initial_lat_count[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,\n"rgw_host", "$1", "ceph_daemon", "rgw.(.*)")',
+ |||
+ label_replace(
+ rate(ceph_rgw_get_initial_lat_sum[30s]) /
+ rate(ceph_rgw_get_initial_lat_count[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,
+ "rgw_host", "$1", "ceph_daemon", "rgw.(.*)"
+ )
+ |||,
'{{rgw_host}}',
15,
1,
'Total bytes transferred in/out through get/put operations, by radosgw instance',
'bytes',
'short',
- 'label_replace(sum by (instance_id) (\n rate(ceph_rgw_get_b[30s]) + \n rate(ceph_rgw_put_b[30s])\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata, "rgw_host", "$1", "ceph_daemon", "rgw.(.*)")',
+ |||
+ label_replace(sum by (instance_id) (
+ rate(ceph_rgw_get_b[30s]) +
+ rate(ceph_rgw_put_b[30s])) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,
+ "rgw_host", "$1", "ceph_daemon", "rgw.(.*)"
+ )
+ |||,
'{{rgw_host}}',
8,
8,
'Latencies are shown stacked, without a yaxis to provide a visual indication of PUT latency imbalance across RGW hosts',
's',
'short',
- 'label_replace(\n rate(ceph_rgw_put_initial_lat_sum[30s]) /\n rate(ceph_rgw_put_initial_lat_count[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,\n"rgw_host", "$1", "ceph_daemon", "rgw.(.*)")',
+ |||
+ label_replace(
+ rate(ceph_rgw_put_initial_lat_sum[30s]) /
+ rate(ceph_rgw_put_initial_lat_count[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,
+ "rgw_host", "$1", "ceph_daemon", "rgw.(.*)"
+ )
+ |||,
'{{rgw_host}}',
15,
8,
'',
'short',
'short',
- 'sum(irate(haproxy_frontend_http_responses_total{code=~"$code",instance=~"$ingress_service",proxy=~"frontend"}[5m])) by (code)',
+ |||
+ sum(
+ irate(
+ haproxy_frontend_http_responses_total{code=~"$code", instance=~"$ingress_service", proxy=~"frontend"}[5m]
+ )
+ ) by (code)
+ |||,
'Frontend {{ code }}',
0,
12,
true
)
.addTargets(
- [u.addTargetSchema('sum(irate(haproxy_backend_http_responses_total{code=~"$code",instance=~"$ingress_service",proxy=~"backend"}[5m])) by (code)', 'Backend {{ code }}')]
+ [
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_backend_http_responses_total{code=~"$code", instance=~"$ingress_service", proxy=~"backend"}[5m]
+ )
+ ) by (code)
+ |||, 'Backend {{ code }}'
+ ),
+ ]
)
.addSeriesOverride([
{
'',
'short',
'short',
- 'sum(irate(haproxy_frontend_http_requests_total{proxy=~"frontend",instance=~"$ingress_service"}[5m])) by (instance)',
+ |||
+ sum(
+ irate(
+ haproxy_frontend_http_requests_total{proxy=~"frontend", instance=~"$ingress_service"}[5m]
+ )
+ ) by (instance)
+ |||,
'Requests',
5,
12,
)
.addTargets(
[
- u.addTargetSchema('sum(irate(haproxy_backend_response_errors_total{proxy=~"backend",instance=~"$ingress_service"}[5m])) by (instance)', 'Response errors', 'time_series', 2),
- u.addTargetSchema('sum(irate(haproxy_frontend_request_errors_total{proxy=~"frontend",instance=~"$ingress_service"}[5m])) by (instance)', 'Requests errors'),
- u.addTargetSchema('sum(irate(haproxy_backend_redispatch_warnings_total{proxy=~"backend",instance=~"$ingress_service"}[5m])) by (instance)', 'Backend redispatch', 'time_series', 2),
- u.addTargetSchema('sum(irate(haproxy_backend_retry_warnings_total{proxy=~"backend",instance=~"$ingress_service"}[5m])) by (instance)', 'Backend retry', 'time_series', 2),
- u.addTargetSchema('sum(irate(haproxy_frontend_requests_denied_total{proxy=~"frontend",instance=~"$ingress_service"}[5m])) by (instance)', 'Request denied', 'time_series', 2),
- u.addTargetSchema('sum(haproxy_backend_current_queue{proxy=~"backend",instance=~"$ingress_service"}) by (instance)', 'Backend Queued', 'time_series', 2),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_backend_response_errors_total{proxy=~"backend",instance=~"$ingress_service"}[5m]
+ )
+ ) by (instance)
+ |||, 'Response errors', 'time_series', 2
+ ),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_frontend_request_errors_total{proxy=~"frontend", instance=~"$ingress_service"}[5m]
+ )
+ ) by (instance)
+ |||, 'Requests errors'
+ ),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_backend_redispatch_warnings_total{proxy=~"backend", instance=~"$ingress_service"}[5m]
+ )
+ ) by (instance)
+ |||, 'Backend redispatch', 'time_series', 2
+ ),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_backend_retry_warnings_total{proxy=~"backend", instance=~"$ingress_service"}[5m]
+ )
+ ) by (instance)
+ |||, 'Backend retry', 'time_series', 2
+ ),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_frontend_requests_denied_total{proxy=~"frontend", instance=~"$ingress_service"}[5m]
+ )
+ ) by (instance)
+ |||, 'Request denied', 'time_series', 2
+ ),
+ u.addTargetSchema(
+ |||
+ sum(
+ haproxy_backend_current_queue{proxy=~"backend", instance=~"$ingress_service"}
+ ) by (instance)
+ |||, 'Backend Queued', 'time_series', 2
+ ),
]
)
.addSeriesOverride([
'',
'short',
'short',
- 'sum(irate(haproxy_frontend_connections_total{proxy=~"frontend",instance=~"$ingress_service"}[5m])) by (instance)',
+ |||
+ sum(
+ irate(
+ haproxy_frontend_connections_total{proxy=~"frontend", instance=~"$ingress_service"}[5m]
+ )
+ ) by (instance)
+ |||,
'Front',
10,
12,
)
.addTargets(
[
- u.addTargetSchema('sum(irate(haproxy_backend_connection_attempts_total{proxy=~"backend",instance=~"$ingress_service"}[5m])) by (instance)', 'Back'),
- u.addTargetSchema('sum(irate(haproxy_backend_connection_errors_total{proxy=~"backend",instance=~"$ingress_service"}[5m])) by (instance)', 'Back errors'),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_backend_connection_attempts_total{proxy=~"backend", instance=~"$ingress_service"}[5m]
+ )
+ ) by (instance)
+ |||, 'Back'
+ ),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_backend_connection_errors_total{proxy=~"backend", instance=~"$ingress_service"}[5m]
+ )
+ ) by (instance)
+ |||, 'Back errors'
+ ),
]
)
.addSeriesOverride([
'',
'short',
'short',
- 'sum(irate(haproxy_frontend_bytes_in_total{proxy=~"frontend",instance=~"$ingress_service"}[5m])*8) by (instance)',
+ |||
+ sum(
+ irate(
+ haproxy_frontend_bytes_in_total{proxy=~"frontend", instance=~"$ingress_service"}[5m]
+ ) * 8
+ ) by (instance)
+ |||,
'IN Front',
15,
12,
)
.addTargets(
[
- u.addTargetSchema('sum(irate(haproxy_frontend_bytes_out_total{proxy=~"frontend",instance=~"$ingress_service"}[5m])*8) by (instance)', 'OUT Front', 'time_series', 2),
- u.addTargetSchema('sum(irate(haproxy_backend_bytes_in_total{proxy=~"backend",instance=~"$ingress_service"}[5m])*8) by (instance)', 'IN Back', 'time_series', 2),
- u.addTargetSchema('sum(irate(haproxy_backend_bytes_out_total{proxy=~"backend",instance=~"$ingress_service"}[5m])*8) by (instance)', 'OUT Back', 'time_series', 2),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_frontend_bytes_out_total{proxy=~"frontend", instance=~"$ingress_service"}[5m]
+ ) * 8
+ ) by (instance)
+ |||, 'OUT Front', 'time_series', 2
+ ),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_backend_bytes_in_total{proxy=~"backend", instance=~"$ingress_service"}[5m]
+ ) * 8
+ ) by (instance)
+ |||, 'IN Back', 'time_series', 2
+ ),
+ u.addTargetSchema(
+ |||
+ sum(
+ irate(
+ haproxy_backend_bytes_out_total{proxy=~"backend", instance=~"$ingress_service"}[5m]
+ ) * 8
+ ) by (instance)
+ |||, 'OUT Back', 'time_series', 2
+ ),
]
)
.addSeriesOverride([
'',
's',
'short',
- 'sum by (instance_id) (rate(ceph_rgw_get_initial_lat_sum[30s]) / rate(ceph_rgw_get_initial_lat_count[30s])) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
- 'sum by (instance_id) (rate(ceph_rgw_put_initial_lat_sum[30s]) / rate(ceph_rgw_put_initial_lat_count[30s])) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
+ |||
+ sum by (instance_id) (
+ rate(ceph_rgw_get_initial_lat_sum[30s]) / rate(ceph_rgw_get_initial_lat_count[30s])
+ ) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
+ |||
+ sum by (instance_id) (
+ rate(ceph_rgw_put_initial_lat_sum[30s]) / rate(ceph_rgw_put_initial_lat_count[30s])
+ ) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
'GET {{ceph_daemon}}',
'PUT {{ceph_daemon}}',
0,
'',
'bytes',
'short',
- 'rate(ceph_rgw_get_b[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
- 'rate(ceph_rgw_put_b[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
+ |||
+ rate(ceph_rgw_get_b[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
+ |||
+ rate(ceph_rgw_put_b[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
'GETs {{ceph_daemon}}',
'PUTs {{ceph_daemon}}',
6,
'',
'short',
'short',
- 'rate(ceph_rgw_failed_req[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
- 'rate(ceph_rgw_get[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
+ |||
+ rate(ceph_rgw_failed_req[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
+ |||
+ rate(ceph_rgw_get[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
'Requests Failed {{ceph_daemon}}',
'GETs {{ceph_daemon}}',
13,
.addTargets(
[
u.addTargetSchema(
- 'rate(ceph_rgw_put[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
+ |||
+ rate(ceph_rgw_put[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
'PUTs {{ceph_daemon}}'
),
u.addTargetSchema(
- '(\n rate(ceph_rgw_req[30s]) -\n (rate(ceph_rgw_get[30s]) + rate(ceph_rgw_put[30s]))\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
+ |||
+ (
+ rate(ceph_rgw_req[30s]) -
+ (rate(ceph_rgw_get[30s]) + rate(ceph_rgw_put[30s]))
+ ) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
'Other {{ceph_daemon}}'
),
]
}, '$datasource', '', 'Under graph', 'pie', 'Workload Breakdown', 'current'
)
.addTarget(u.addTargetSchema(
- 'rate(ceph_rgw_failed_req[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
+ |||
+ rate(ceph_rgw_failed_req[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
'Failures {{ceph_daemon}}'
))
.addTarget(u.addTargetSchema(
- 'rate(ceph_rgw_get[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
+ |||
+ rate(ceph_rgw_get[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
'GETs {{ceph_daemon}}'
))
.addTarget(u.addTargetSchema(
- 'rate(ceph_rgw_put[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
+ |||
+ rate(ceph_rgw_put[30s]) *
+ on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
'PUTs {{ceph_daemon}}'
))
.addTarget(u.addTargetSchema(
- '(\n rate(ceph_rgw_req[30s]) -\n (rate(ceph_rgw_get[30s]) + rate(ceph_rgw_put[30s]))\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}',
+ |||
+ (
+ rate(ceph_rgw_req[30s]) -
+ (rate(ceph_rgw_get[30s]) + rate(ceph_rgw_put[30s]))
+ ) * on (instance_id) group_left (ceph_daemon)
+ ceph_rgw_metadata{ceph_daemon=~"$rgw_servers"}
+ |||,
'Other (DELETE,LIST) {{ceph_daemon}}'
)) + { gridPos: { x: 20, y: 1, w: 4, h: 8 } },
]),
"steppedLine": false,
"targets": [
{
- "expr": "sum by (mode) (\n irate(node_cpu{instance=~\"($ceph_hosts)([\\\\.:].*)?\", mode=~\"(irq|nice|softirq|steal|system|user|iowait)\"}[1m]) or\n irate(node_cpu_seconds_total{instance=~\"($ceph_hosts)([\\\\.:].*)?\", mode=~\"(irq|nice|softirq|steal|system|user|iowait)\"}[1m])\n) / scalar(\n sum(irate(node_cpu{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[1m]) or\n irate(node_cpu_seconds_total{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[1m]))\n) * 100",
+ "expr": "sum by (mode) (\n irate(node_cpu{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\", mode=~\"(irq|nice|softirq|steal|system|user|iowait)\"}[1m]) or\n irate(node_cpu_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\", mode=~\"(irq|nice|softirq|steal|system|user|iowait)\"}[1m])\n) / (\n scalar(\n sum(irate(node_cpu{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[1m]) or\n irate(node_cpu_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[1m]))\n ) * 100\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{mode}}",
"steppedLine": false,
"targets": [
{
- "expr": "node_memory_MemFree{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_MemFree_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"} ",
+ "expr": "node_memory_MemFree{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_MemFree_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Free",
"refId": "A"
},
{
- "expr": "node_memory_MemTotal{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_MemTotal_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"} ",
+ "expr": "node_memory_MemTotal{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_MemTotal_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "total",
"refId": "B"
},
{
- "expr": "(node_memory_Cached{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_Cached_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"}) + \n(node_memory_Buffers{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_Buffers_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"}) +\n(node_memory_Slab{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_Slab_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"}) \n",
+ "expr": "(\n node_memory_Cached{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Cached_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n) + (\n node_memory_Buffers{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Buffers_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n) + (\n node_memory_Slab{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Slab_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "buffers/cache",
"refId": "C"
},
{
- "expr": "(node_memory_MemTotal{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_MemTotal_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"})- (\n (node_memory_MemFree{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_MemFree_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"}) + \n (node_memory_Cached{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_Cached_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"}) + \n (node_memory_Buffers{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_Buffers_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"}) +\n (node_memory_Slab{instance=~\"$ceph_hosts([\\\\.:].*)?\"} or node_memory_Slab_bytes{instance=~\"$ceph_hosts([\\\\.:].*)?\"})\n )\n \n",
+ "expr": "(\n node_memory_MemTotal{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_MemTotal_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n) - (\n (\n node_memory_MemFree{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_MemFree_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n ) + (\n node_memory_Cached{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Cached_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n ) + (\n node_memory_Buffers{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Buffers_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n ) +\n (\n node_memory_Slab{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Slab_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n )\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "used",
"steppedLine": false,
"targets": [
{
- "expr": "sum by (device) (\n irate(node_network_receive_bytes{instance=~\"($ceph_hosts)([\\\\.:].*)?\",device!=\"lo\"}[1m]) or \n irate(node_network_receive_bytes_total{instance=~\"($ceph_hosts)([\\\\.:].*)?\",device!=\"lo\"}[1m])\n)",
+ "expr": "sum by (device) (\n irate(\n node_network_receive_bytes{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\",device!=\"lo\"}[1m]) or\n irate(node_network_receive_bytes_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\",device!=\"lo\"}[1m]\n )\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}.rx",
"refId": "A"
},
{
- "expr": "sum by (device) (\n irate(node_network_transmit_bytes{instance=~\"($ceph_hosts)([\\\\.:].*)?\",device!=\"lo\"}[1m]) or\n irate(node_network_transmit_bytes_total{instance=~\"($ceph_hosts)([\\\\.:].*)?\",device!=\"lo\"}[1m])\n)",
+ "expr": "sum by (device) (\n irate(node_network_transmit_bytes{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\",device!=\"lo\"}[1m]) or\n irate(node_network_transmit_bytes_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\",device!=\"lo\"}[1m])\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}.tx",
"steppedLine": false,
"targets": [
{
- "expr": "irate(node_network_receive_drop{instance=~\"$ceph_hosts([\\\\.:].*)?\"}[1m]) or irate(node_network_receive_drop_total{instance=~\"$ceph_hosts([\\\\.:].*)?\"}[1m])",
+ "expr": "irate(node_network_receive_drop{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[1m]) or\n irate(node_network_receive_drop_total{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[1m])\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}.rx",
"refId": "A"
},
{
- "expr": "irate(node_network_transmit_drop{instance=~\"$ceph_hosts([\\\\.:].*)?\"}[1m]) or irate(node_network_transmit_drop_total{instance=~\"$ceph_hosts([\\\\.:].*)?\"}[1m])",
+ "expr": "irate(node_network_transmit_drop{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[1m]) or\n irate(node_network_transmit_drop_total{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[1m])\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}.tx",
"tableColumn": "",
"targets": [
{
- "expr": "sum(ceph_osd_stat_bytes and on (ceph_daemon) ceph_disk_occupation{instance=~\"($ceph_hosts)([\\\\.:].*)?\"})",
+ "expr": "sum(\n ceph_osd_stat_bytes and\n on (ceph_daemon) ceph_disk_occupation{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "",
"steppedLine": false,
"targets": [
{
- "expr": "irate(node_network_receive_errs{instance=~\"$ceph_hosts([\\\\.:].*)?\"}[1m]) or irate(node_network_receive_errs_total{instance=~\"$ceph_hosts([\\\\.:].*)?\"}[1m])",
+ "expr": "irate(node_network_receive_errs{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[1m]) or\n irate(node_network_receive_errs_total{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[1m])\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}.rx",
"refId": "A"
},
{
- "expr": "irate(node_network_transmit_errs{instance=~\"$ceph_hosts([\\\\.:].*)?\"}[1m]) or irate(node_network_transmit_errs_total{instance=~\"$ceph_hosts([\\\\.:].*)?\"}[1m])",
+ "expr": "irate(node_network_transmit_errs{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[1m]) or\n irate(node_network_transmit_errs_total{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[1m])\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}.tx",
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(\n (\n irate(node_disk_writes_completed{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]) or\n irate(node_disk_writes_completed_total{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m])\n ),\n \"instance\",\n \"$1\",\n \"instance\",\n \"([^:.]*).*\"\n)\n* on(instance, device) group_left(ceph_daemon)\n label_replace(\n label_replace(\n ceph_disk_occupation_human,\n \"device\",\n \"$1\",\n \"device\",\n \"/dev/(.*)\"\n ),\n \"instance\",\n \"$1\",\n \"instance\",\n \"([^:.]*).*\"\n )",
+ "expr": "label_replace(\n (\n irate(node_disk_writes_completed{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m]) or\n irate(node_disk_writes_completed_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m])\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(\n ceph_disk_occupation_human, \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}({{ceph_daemon}}) writes",
"refId": "A"
},
{
- "expr": "label_replace(\n (irate(node_disk_reads_completed{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]) or irate(node_disk_reads_completed_total{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m])),\n \"instance\",\n \"$1\",\n \"instance\",\n \"([^:.]*).*\"\n)\n* on(instance, device) group_left(ceph_daemon)\n label_replace(\n label_replace(\n ceph_disk_occupation_human,\n \"device\",\n \"$1\",\n \"device\",\n \"/dev/(.*)\"\n ),\n \"instance\",\n \"$1\",\n \"instance\",\n \"([^:.]*).*\"\n )",
+ "expr": "label_replace(\n (\n irate(node_disk_reads_completed{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m]) or\n irate(node_disk_reads_completed_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m])\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(\n ceph_disk_occupation_human,\"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}({{ceph_daemon}}) reads",
"steppedLine": false,
"targets": [
{
- "expr": "label_replace((irate(node_disk_bytes_written{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]) or irate(node_disk_written_bytes_total{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m])), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")",
+ "expr": "label_replace(\n (\n irate(node_disk_bytes_written{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m]) or\n irate(node_disk_written_bytes_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m])\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") * on(instance, device)\n group_left(ceph_daemon) label_replace(\n label_replace(ceph_disk_occupation_human, \"device\", \"$1\", \"device\", \"/dev/(.*)\"),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n )\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}({{ceph_daemon}}) write",
"refId": "A"
},
{
- "expr": "label_replace((irate(node_disk_bytes_read{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]) or irate(node_disk_read_bytes_total{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m])), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")",
+ "expr": "label_replace(\n (\n irate(node_disk_bytes_read{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m]) or\n irate(node_disk_read_bytes_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m])\n ),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") * on(instance, device)\n group_left(ceph_daemon) label_replace(\n label_replace(ceph_disk_occupation_human, \"device\", \"$1\", \"device\", \"/dev/(.*)\"),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n )\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}({{ceph_daemon}}) read",
"steppedLine": false,
"targets": [
{
- "expr": "max by(instance,device) (label_replace((irate(node_disk_write_time_seconds_total{ instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]) ) / clamp_min(irate(node_disk_writes_completed_total{ instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]), 0.001) or (irate(node_disk_read_time_seconds_total{ instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]) ) / clamp_min(irate(node_disk_reads_completed_total{ instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]), 0.001), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")",
+ "expr": "max by(instance, device) (label_replace(\n (irate(node_disk_write_time_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m])) /\n clamp_min(irate(node_disk_writes_completed_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m]), 0.001) or\n (irate(node_disk_read_time_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m])) /\n clamp_min(irate(node_disk_reads_completed_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m]), 0.001),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(\n ceph_disk_occupation_human{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}({{ceph_daemon}})",
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(((irate(node_disk_io_time_ms{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]) / 10 ) or irate(node_disk_io_time_seconds_total{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}[5m]) * 100), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human{instance=~\"($ceph_hosts)([\\\\.:].*)?\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")",
+ "expr": "label_replace(\n (\n (irate(node_disk_io_time_ms{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m]) / 10) or\n irate(node_disk_io_time_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[5m]) * 100\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(ceph_disk_occupation_human{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}}({{ceph_daemon}})",
"tableColumn": "",
"targets": [
{
- "expr": "avg(\n 1 - (\n avg by(instance) \n (irate(node_cpu_seconds_total{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]) or\n irate(node_cpu{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]))\n )\n )",
+ "expr": "avg(1 - (\n avg by(instance) (\n irate(node_cpu_seconds_total{mode=\\'idle\\',instance=~\\\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\\\"}[1m]) or\n irate(node_cpu{mode=\\'idle\\',instance=~\\\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\\\"}[1m])\n )\n))\n",
"format": "time_series",
"instant": true,
"intervalFactor": 1,
"tableColumn": "",
"targets": [
{
- "expr": "avg (((node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_MemTotal_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"})- (\n (node_memory_MemFree{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_MemFree_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}) + \n (node_memory_Cached{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_Cached_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}) + \n (node_memory_Buffers{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_Buffers_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}) +\n (node_memory_Slab{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_Slab_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"})\n )) /\n (node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_MemTotal_bytes{instance=~\"($osd_hosts|$rgw_hosts|$mon_hosts|$mds_hosts).*\"} ))",
+ "expr": "avg ((\n (\n node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_MemTotal_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}\n ) - ((\n node_memory_MemFree{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_MemFree_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}) +\n (\n node_memory_Cached{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_Cached_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}\n ) + (\n node_memory_Buffers{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_Buffers_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}\n ) + (\n node_memory_Slab{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_Slab_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}\n )\n )\n)\n(\n node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_MemTotal_bytes{instance=~\"($osd_hosts|$rgw_hosts|$mon_hosts|$mds_hosts).*\"}\n))\n",
"format": "time_series",
"instant": true,
"intervalFactor": 1,
"tableColumn": "",
"targets": [
{
- "expr": "sum ((irate(node_disk_reads_completed{instance=~\"($osd_hosts).*\"}[5m]) or irate(node_disk_reads_completed_total{instance=~\"($osd_hosts).*\"}[5m]) ) + \n(irate(node_disk_writes_completed{instance=~\"($osd_hosts).*\"}[5m]) or irate(node_disk_writes_completed_total{instance=~\"($osd_hosts).*\"}[5m])))",
+ "expr": "sum ((\n irate(node_disk_reads_completed{instance=~\"($osd_hosts).*\"}[5m]) or\n irate(node_disk_reads_completed_total{instance=~\"($osd_hosts).*\"}[5m])\n) + (\n irate(node_disk_writes_completed{instance=~\"($osd_hosts).*\"}[5m]) or\n irate(node_disk_writes_completed_total{instance=~\"($osd_hosts).*\"}[5m])\n))\n",
"format": "time_series",
"instant": true,
"intervalFactor": 1,
"tableColumn": "",
"targets": [
{
- "expr": "avg (\n label_replace((irate(node_disk_io_time_ms[5m]) / 10 ) or\n (irate(node_disk_io_time_seconds_total[5m]) * 100), \"instance\", \"$1\", \"instance\", \"([^.:]*).*\"\n ) *\n on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation_human{instance=~\"($osd_hosts).*\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^.:]*).*\")\n)",
+ "expr": "avg (\n label_replace(\n (irate(node_disk_io_time_ms[5m]) / 10 ) or\n (irate(node_disk_io_time_seconds_total[5m]) * 100),\n \"instance\", \"$1\", \"instance\", \"([^.:]*).*\"\n ) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(\n ceph_disk_occupation_human{instance=~\"($osd_hosts).*\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^.:]*).*\"\n )\n)\n",
"format": "time_series",
"instant": true,
"intervalFactor": 1,
"steppedLine": false,
"targets": [
{
- "expr": "topk(10,100 * ( 1 - (\n avg by(instance) \n (irate(node_cpu_seconds_total{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]) or\n irate(node_cpu{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]))\n )\n )\n)",
+ "expr": "topk(10,\n 100 * (\n 1 - (\n avg by(instance) (\n irate(node_cpu_seconds_total{mode=\\'idle\\',instance=~\\\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\\\"}[1m]) or\n irate(node_cpu{mode=\\'idle\\',instance=~\\\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\\\"}[1m])\n )\n )\n )\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{instance}}",
"steppedLine": false,
"targets": [
{
- "expr": "irate(ceph_osd_op_r_latency_sum{ceph_daemon=~\"$osd\"}[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m])",
+ "expr": "irate(ceph_osd_op_r_latency_sum{ceph_daemon=~\"$osd\"}[1m]) /\n on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m])\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "read",
"refId": "A"
},
{
- "expr": "irate(ceph_osd_op_w_latency_sum{ceph_daemon=~\"$osd\"}[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m])",
+ "expr": "irate(ceph_osd_op_w_latency_sum{ceph_daemon=~\"$osd\"}[1m]) /\n on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m])\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "write",
"steppedLine": false,
"targets": [
{
- "expr": "(label_replace(irate(node_disk_read_time_seconds_total[1m]) / irate(node_disk_reads_completed_total[1m]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"))",
+ "expr": "(\n label_replace(\n irate(node_disk_read_time_seconds_total[1m]) / irate(node_disk_reads_completed_total[1m]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n ) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n )\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{instance}}/{{device}} Reads",
"refId": "A"
},
{
- "expr": "(label_replace(irate(node_disk_write_time_seconds_total[1m]) / irate(node_disk_writes_completed_total[1m]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"))",
+ "expr": "(\n label_replace(\n irate(node_disk_write_time_seconds_total[1m]) / irate(node_disk_writes_completed_total[1m]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") and on (instance, device)\n label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n )\n )\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{instance}}/{{device}} Writes",
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(irate(node_disk_writes_completed_total[1m]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")",
+ "expr": "label_replace(\n irate(node_disk_writes_completed_total[1m]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}} on {{instance}} Writes",
"refId": "A"
},
{
- "expr": "label_replace(irate(node_disk_reads_completed_total[1m]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")",
+ "expr": "label_replace(\n irate(node_disk_reads_completed_total[1m]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}} on {{instance}} Reads",
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(irate(node_disk_read_bytes_total[1m]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")",
+ "expr": "label_replace(\n irate(node_disk_read_bytes_total[1m]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{instance}} {{device}} Reads",
"refId": "A"
},
{
- "expr": "label_replace(irate(node_disk_written_bytes_total[1m]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")",
+ "expr": "label_replace(\n irate(node_disk_written_bytes_total[1m]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{instance}} {{device}} Writes",
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(irate(node_disk_io_time_seconds_total[1m]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") and on (instance, device) label_replace(label_replace(ceph_disk_occupation_human{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\")",
+ "expr": "label_replace(\n irate(node_disk_io_time_seconds_total[1m]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{device}} on {{instance}}",
"refId": "A"
},
{
- "expr": "max (irate(ceph_osd_op_r_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000)",
+ "expr": "max(\n irate(ceph_osd_op_r_latency_sum[1m]) /\n on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "MAX read",
"refId": "B"
},
{
- "expr": "quantile(0.95,\n (irate(ceph_osd_op_r_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000)\n)",
+ "expr": "quantile(0.95,\n (\n irate(ceph_osd_op_r_latency_sum[1m]) /\n on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m])\n * 1000\n )\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "@95%ile",
],
"targets": [
{
- "expr": "topk(10,\n (sort(\n (irate(ceph_osd_op_r_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000)\n ))\n)\n\n",
+ "expr": "topk(10,\n (sort(\n (\n irate(ceph_osd_op_r_latency_sum[1m]) /\n on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) *\n 1000\n )\n ))\n)\n",
"format": "table",
"instant": true,
"intervalFactor": 1,
"steppedLine": false,
"targets": [
{
- "expr": "avg (irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000)",
+ "expr": "avg(\n irate(ceph_osd_op_w_latency_sum[1m]) /\n on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m])\n * 1000\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "AVG write",
"refId": "A"
},
{
- "expr": "max (irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000)",
+ "expr": "max(\n irate(ceph_osd_op_w_latency_sum[1m]) /\n on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) *\n 1000\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "MAX write",
"refId": "B"
},
{
- "expr": "quantile(0.95,\n (irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000)\n)",
+ "expr": "quantile(0.95, (\n irate(ceph_osd_op_w_latency_sum[1m]) /\n on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) *\n 1000\n))\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "@95%ile write",
],
"targets": [
{
- "expr": "topk(10,\n (sort(\n (irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000)\n ))\n)\n\n",
+ "expr": "topk(10,\n (sort(\n (irate(ceph_osd_op_w_latency_sum[1m]) /\n on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) *\n 1000)\n ))\n)\n",
"format": "table",
"instant": true,
"intervalFactor": 1,
"steppedLine": false,
"targets": [
{
- "expr": "ceph_osd_numpg\n",
+ "expr": "ceph_osd_numpg",
"format": "time_series",
"instant": true,
"intervalFactor": 1,
"tableColumn": "",
"targets": [
{
- "expr": "sum(ceph_bluestore_onode_hits)/(sum(ceph_bluestore_onode_hits) + sum(ceph_bluestore_onode_misses))",
+ "expr": "sum(ceph_bluestore_onode_hits) / (\n sum(ceph_bluestore_onode_hits) +\n sum(ceph_bluestore_onode_misses)\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "",
"tableColumn": "",
"targets": [
{
- "expr": "(ceph_pool_stored / (ceph_pool_stored + ceph_pool_max_avail)) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}",
+ "expr": "(ceph_pool_stored / (ceph_pool_stored + ceph_pool_max_avail)) *\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "",
"tableColumn": "",
"targets": [
{
- "expr": "(ceph_pool_max_avail / deriv(ceph_pool_stored[6h])) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"} > 0",
+ "expr": "(ceph_pool_max_avail / deriv(ceph_pool_stored[6h])) *\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"} > 0\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "",
"steppedLine": false,
"targets": [
{
- "expr": "deriv(ceph_pool_objects[1m]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}",
+ "expr": "deriv(ceph_pool_objects[1m]) *\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Objects per second",
"steppedLine": false,
"targets": [
{
- "expr": "irate(ceph_pool_rd[1m]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}",
+ "expr": "irate(ceph_pool_rd[1m]) *\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "reads",
"refId": "A"
},
{
- "expr": "irate(ceph_pool_wr[1m]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}",
+ "expr": "irate(ceph_pool_wr[1m]) *\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "writes",
"steppedLine": false,
"targets": [
{
- "expr": "irate(ceph_pool_rd_bytes[1m]) + on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}",
+ "expr": "irate(ceph_pool_rd_bytes[1m]) +\n on(pool_id) group_left(instance, name) ceph_pool_metadata{name=~\"$pool_name\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "reads",
"refId": "A"
},
{
- "expr": "irate(ceph_pool_wr_bytes[1m]) + on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}",
+ "expr": "irate(ceph_pool_wr_bytes[1m]) +\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "writes",
"steppedLine": false,
"targets": [
{
- "expr": "ceph_pool_objects * on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}",
+ "expr": "ceph_pool_objects *\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Number of Objects",
"#d44a3a"
],
"datasource": "$datasource",
- "description": "Indicates how suitable the data is within the pools that are/have been enabled for compression - averaged across all pools holding compressed data\n",
+ "description": "Indicates how suitable the data is within the pools that are/have been enabled for compression - averaged across all pools holding compressed data",
"format": "percent",
"gauge": {
"maxValue": 100,
"tableColumn": "",
"targets": [
{
- "expr": "(sum(ceph_pool_compress_under_bytes > 0) / sum(ceph_pool_stored_raw and ceph_pool_compress_under_bytes > 0)) * 100",
+ "expr": "(\n sum(ceph_pool_compress_under_bytes > 0) /\n sum(ceph_pool_stored_raw and ceph_pool_compress_under_bytes > 0)\n) * 100\n",
"format": "table",
"intervalFactor": 1,
"legendFormat": "",
],
"targets": [
{
- "expr": "(ceph_pool_compress_under_bytes / ceph_pool_compress_bytes_used > 0) and on(pool_id) (((ceph_pool_compress_under_bytes > 0) / ceph_pool_stored_raw) * 100 > 0.5)",
+ "expr": "(ceph_pool_compress_under_bytes / ceph_pool_compress_bytes_used > 0) and on(pool_id) (\n ((ceph_pool_compress_under_bytes > 0) / ceph_pool_stored_raw) * 100 > 0.5\n)\n",
"format": "table",
"instant": true,
"intervalFactor": 1,
"steppedLine": false,
"targets": [
{
- "expr": "topk($topk,round((rate(ceph_pool_rd[30s]) + rate(ceph_pool_wr[30s])),1) * on(pool_id) group_left(instance,name) ceph_pool_metadata) ",
+ "expr": "topk($topk,\n round(\n (rate(ceph_pool_rd[30s]) + rate(ceph_pool_wr[30s])),\n 1\n ) * on(pool_id) group_left(instance,name) ceph_pool_metadata)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{name}} ",
"refId": "A"
},
{
- "expr": "topk($topk,rate(ceph_pool_wr[30s]) + on(pool_id) group_left(instance,name) ceph_pool_metadata) ",
+ "expr": "topk($topk,\n rate(ceph_pool_wr[30s]) +\n on(pool_id) group_left(instance,name) ceph_pool_metadata\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{name}} - write",
"steppedLine": false,
"targets": [
{
- "expr": "topk($topk,(rate(ceph_pool_rd_bytes[30s]) + rate(ceph_pool_wr_bytes[30s])) * on(pool_id) group_left(instance,name) ceph_pool_metadata)",
+ "expr": "topk($topk,\n (rate(ceph_pool_rd_bytes[30s]) + rate(ceph_pool_wr_bytes[30s])) *\n on(pool_id) group_left(instance, name) ceph_pool_metadata\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{name}}",
"steppedLine": false,
"targets": [
{
- "expr": "sum by (instance_id) (rate(ceph_rgw_get_initial_lat_sum[30s]) / rate(ceph_rgw_get_initial_lat_count[30s])) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "sum by (instance_id) (\n rate(ceph_rgw_get_initial_lat_sum[30s]) / rate(ceph_rgw_get_initial_lat_count[30s])\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "GET {{ceph_daemon}}",
"refId": "A"
},
{
- "expr": "sum by (instance_id) (rate(ceph_rgw_put_initial_lat_sum[30s]) / rate(ceph_rgw_put_initial_lat_count[30s])) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "sum by (instance_id) (\n rate(ceph_rgw_put_initial_lat_sum[30s]) / rate(ceph_rgw_put_initial_lat_count[30s])\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUT {{ceph_daemon}}",
"steppedLine": false,
"targets": [
{
- "expr": "rate(ceph_rgw_get_b[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "rate(ceph_rgw_get_b[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "GETs {{ceph_daemon}}",
"refId": "A"
},
{
- "expr": "rate(ceph_rgw_put_b[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "rate(ceph_rgw_put_b[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUTs {{ceph_daemon}}",
"steppedLine": false,
"targets": [
{
- "expr": "rate(ceph_rgw_failed_req[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "rate(ceph_rgw_failed_req[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Requests Failed {{ceph_daemon}}",
"refId": "A"
},
{
- "expr": "rate(ceph_rgw_get[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "rate(ceph_rgw_get[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "GETs {{ceph_daemon}}",
"refId": "B"
},
{
- "expr": "rate(ceph_rgw_put[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "rate(ceph_rgw_put[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUTs {{ceph_daemon}}",
"refId": "C"
},
{
- "expr": "(\n rate(ceph_rgw_req[30s]) -\n (rate(ceph_rgw_get[30s]) + rate(ceph_rgw_put[30s]))\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "(\n rate(ceph_rgw_req[30s]) -\n (rate(ceph_rgw_get[30s]) + rate(ceph_rgw_put[30s]))\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Other {{ceph_daemon}}",
"pieType": "pie",
"targets": [
{
- "expr": "rate(ceph_rgw_failed_req[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "rate(ceph_rgw_failed_req[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Failures {{ceph_daemon}}",
"refId": "A"
},
{
- "expr": "rate(ceph_rgw_get[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "rate(ceph_rgw_get[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "GETs {{ceph_daemon}}",
"refId": "B"
},
{
- "expr": "rate(ceph_rgw_put[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "rate(ceph_rgw_put[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUTs {{ceph_daemon}}",
"refId": "C"
},
{
- "expr": "(\n rate(ceph_rgw_req[30s]) -\n (rate(ceph_rgw_get[30s]) + rate(ceph_rgw_put[30s]))\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}",
+ "expr": "(\n rate(ceph_rgw_req[30s]) -\n (rate(ceph_rgw_get[30s]) + rate(ceph_rgw_put[30s]))\n) * on (instance_id) group_left (ceph_daemon)\n ceph_rgw_metadata{ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Other (DELETE,LIST) {{ceph_daemon}}",
"steppedLine": false,
"targets": [
{
- "expr": "rate(ceph_rgw_get_initial_lat_sum[30s]) / rate(ceph_rgw_get_initial_lat_count[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata",
+ "expr": "rate(ceph_rgw_get_initial_lat_sum[30s]) /\n rate(ceph_rgw_get_initial_lat_count[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "GET AVG",
"refId": "A"
},
{
- "expr": "rate(ceph_rgw_put_initial_lat_sum[30s]) / rate(ceph_rgw_put_initial_lat_count[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata",
+ "expr": "rate(ceph_rgw_put_initial_lat_sum[30s]) /\n rate(ceph_rgw_put_initial_lat_count[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUT AVG",
"steppedLine": false,
"targets": [
{
- "expr": "sum by (rgw_host) (label_replace(rate(ceph_rgw_req[30s]) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata, \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"))",
+ "expr": "sum by (rgw_host) (\n label_replace(\n rate(ceph_rgw_req[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n )\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{rgw_host}}",
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(\n rate(ceph_rgw_get_initial_lat_sum[30s]) /\n rate(ceph_rgw_get_initial_lat_count[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,\n\"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\")",
+ "expr": "label_replace(\n rate(ceph_rgw_get_initial_lat_sum[30s]) /\n rate(ceph_rgw_get_initial_lat_count[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{rgw_host}}",
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(sum by (instance_id) (\n rate(ceph_rgw_get_b[30s]) + \n rate(ceph_rgw_put_b[30s])\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata, \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\")",
+ "expr": "label_replace(sum by (instance_id) (\n rate(ceph_rgw_get_b[30s]) +\n rate(ceph_rgw_put_b[30s])) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{rgw_host}}",
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(\n rate(ceph_rgw_put_initial_lat_sum[30s]) /\n rate(ceph_rgw_put_initial_lat_count[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,\n\"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\")",
+ "expr": "label_replace(\n rate(ceph_rgw_put_initial_lat_sum[30s]) /\n rate(ceph_rgw_put_initial_lat_count[30s]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata,\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{rgw_host}}",
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(haproxy_frontend_http_responses_total{code=~\"$code\",instance=~\"$ingress_service\",proxy=~\"frontend\"}[5m])) by (code)",
+ "expr": "sum(\n irate(\n haproxy_frontend_http_responses_total{code=~\"$code\", instance=~\"$ingress_service\", proxy=~\"frontend\"}[5m]\n )\n) by (code)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Frontend {{ code }}",
"refId": "A"
},
{
- "expr": "sum(irate(haproxy_backend_http_responses_total{code=~\"$code\",instance=~\"$ingress_service\",proxy=~\"backend\"}[5m])) by (code)",
+ "expr": "sum(\n irate(\n haproxy_backend_http_responses_total{code=~\"$code\", instance=~\"$ingress_service\", proxy=~\"backend\"}[5m]\n )\n) by (code)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Backend {{ code }}",
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(haproxy_frontend_http_requests_total{proxy=~\"frontend\",instance=~\"$ingress_service\"}[5m])) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_frontend_http_requests_total{proxy=~\"frontend\", instance=~\"$ingress_service\"}[5m]\n )\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Requests",
"refId": "A"
},
{
- "expr": "sum(irate(haproxy_backend_response_errors_total{proxy=~\"backend\",instance=~\"$ingress_service\"}[5m])) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_backend_response_errors_total{proxy=~\"backend\",instance=~\"$ingress_service\"}[5m]\n )\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "Response errors",
"refId": "B"
},
{
- "expr": "sum(irate(haproxy_frontend_request_errors_total{proxy=~\"frontend\",instance=~\"$ingress_service\"}[5m])) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_frontend_request_errors_total{proxy=~\"frontend\", instance=~\"$ingress_service\"}[5m]\n )\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Requests errors",
"refId": "C"
},
{
- "expr": "sum(irate(haproxy_backend_redispatch_warnings_total{proxy=~\"backend\",instance=~\"$ingress_service\"}[5m])) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_backend_redispatch_warnings_total{proxy=~\"backend\", instance=~\"$ingress_service\"}[5m]\n )\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "Backend redispatch",
"refId": "D"
},
{
- "expr": "sum(irate(haproxy_backend_retry_warnings_total{proxy=~\"backend\",instance=~\"$ingress_service\"}[5m])) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_backend_retry_warnings_total{proxy=~\"backend\", instance=~\"$ingress_service\"}[5m]\n )\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "Backend retry",
"refId": "E"
},
{
- "expr": "sum(irate(haproxy_frontend_requests_denied_total{proxy=~\"frontend\",instance=~\"$ingress_service\"}[5m])) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_frontend_requests_denied_total{proxy=~\"frontend\", instance=~\"$ingress_service\"}[5m]\n )\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "Request denied",
"refId": "F"
},
{
- "expr": "sum(haproxy_backend_current_queue{proxy=~\"backend\",instance=~\"$ingress_service\"}) by (instance)",
+ "expr": "sum(\n haproxy_backend_current_queue{proxy=~\"backend\", instance=~\"$ingress_service\"}\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "Backend Queued",
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(haproxy_frontend_connections_total{proxy=~\"frontend\",instance=~\"$ingress_service\"}[5m])) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_frontend_connections_total{proxy=~\"frontend\", instance=~\"$ingress_service\"}[5m]\n )\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Front",
"refId": "A"
},
{
- "expr": "sum(irate(haproxy_backend_connection_attempts_total{proxy=~\"backend\",instance=~\"$ingress_service\"}[5m])) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_backend_connection_attempts_total{proxy=~\"backend\", instance=~\"$ingress_service\"}[5m]\n )\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Back",
"refId": "B"
},
{
- "expr": "sum(irate(haproxy_backend_connection_errors_total{proxy=~\"backend\",instance=~\"$ingress_service\"}[5m])) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_backend_connection_errors_total{proxy=~\"backend\", instance=~\"$ingress_service\"}[5m]\n )\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Back errors",
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(haproxy_frontend_bytes_in_total{proxy=~\"frontend\",instance=~\"$ingress_service\"}[5m])*8) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_frontend_bytes_in_total{proxy=~\"frontend\", instance=~\"$ingress_service\"}[5m]\n ) * 8\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "IN Front",
"refId": "A"
},
{
- "expr": "sum(irate(haproxy_frontend_bytes_out_total{proxy=~\"frontend\",instance=~\"$ingress_service\"}[5m])*8) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_frontend_bytes_out_total{proxy=~\"frontend\", instance=~\"$ingress_service\"}[5m]\n ) * 8\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "OUT Front",
"refId": "B"
},
{
- "expr": "sum(irate(haproxy_backend_bytes_in_total{proxy=~\"backend\",instance=~\"$ingress_service\"}[5m])*8) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_backend_bytes_in_total{proxy=~\"backend\", instance=~\"$ingress_service\"}[5m]\n ) * 8\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "IN Back",
"refId": "C"
},
{
- "expr": "sum(irate(haproxy_backend_bytes_out_total{proxy=~\"backend\",instance=~\"$ingress_service\"}[5m])*8) by (instance)",
+ "expr": "sum(\n irate(\n haproxy_backend_bytes_out_total{proxy=~\"backend\", instance=~\"$ingress_service\"}[5m]\n ) * 8\n) by (instance)\n",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "OUT Back",
"steppedLine": false,
"targets": [
{
- "expr": "irate(ceph_rbd_write_latency_sum{pool=\"$Pool\", image=\"$Image\"}[30s]) / irate(ceph_rbd_write_latency_count{pool=\"$Pool\", image=\"$Image\"}[30s])",
+ "expr": "irate(ceph_rbd_write_latency_sum{pool=\"$Pool\", image=\"$Image\"}[30s]) /\n irate(ceph_rbd_write_latency_count{pool=\"$Pool\", image=\"$Image\"}[30s])\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{pool}} Write",
"refId": "A"
},
{
- "expr": "irate(ceph_rbd_read_latency_sum{pool=\"$Pool\", image=\"$Image\"}[30s]) / irate(ceph_rbd_read_latency_count{pool=\"$Pool\", image=\"$Image\"}[30s])",
+ "expr": "irate(ceph_rbd_read_latency_sum{pool=\"$Pool\", image=\"$Image\"}[30s]) /\n irate(ceph_rbd_read_latency_count{pool=\"$Pool\", image=\"$Image\"}[30s])\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{pool}} Read",
"steppedLine": false,
"targets": [
{
- "expr": "round(sum(irate(ceph_rbd_write_latency_sum[30s])) / sum(irate(ceph_rbd_write_latency_count[30s])))",
+ "expr": "round(\n sum(irate(ceph_rbd_write_latency_sum[30s])) /\n sum(irate(ceph_rbd_write_latency_count[30s]))\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Write",
"refId": "A"
},
{
- "expr": "round(sum(irate(ceph_rbd_read_latency_sum[30s])) / sum(irate(ceph_rbd_read_latency_count[30s])))",
+ "expr": "round(\n sum(irate(ceph_rbd_read_latency_sum[30s])) /\n sum(irate(ceph_rbd_read_latency_count[30s]))\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "Read",
],
"targets": [
{
- "expr": "topk(10, (sort((irate(ceph_rbd_write_ops[30s]) + on (image, pool, namespace) irate(ceph_rbd_read_ops[30s])))))",
+ "expr": "topk(10,\n (\n sort((\n irate(ceph_rbd_write_ops[30s]) +\n on (image, pool, namespace) irate(ceph_rbd_read_ops[30s])\n ))\n )\n)\n",
"format": "table",
"instant": true,
"intervalFactor": 1,
],
"targets": [
{
- "expr": "topk(10, sort(sum(irate(ceph_rbd_read_bytes[30s]) + irate(ceph_rbd_write_bytes[30s])) by (pool, image, namespace)))",
+ "expr": "topk(10,\n sort(\n sum(\n irate(ceph_rbd_read_bytes[30s]) + irate(ceph_rbd_write_bytes[30s])\n ) by (pool, image, namespace)\n )\n)\n",
"format": "table",
"instant": true,
"intervalFactor": 1,
],
"targets": [
{
- "expr": "topk(10,\n sum(\n irate(ceph_rbd_write_latency_sum[30s]) / clamp_min(irate(ceph_rbd_write_latency_count[30s]), 1) +\n irate(ceph_rbd_read_latency_sum[30s]) / clamp_min(irate(ceph_rbd_read_latency_count[30s]), 1)\n ) by (pool, image, namespace)\n)",
+ "expr": "topk(10,\n sum(\n irate(ceph_rbd_write_latency_sum[30s]) / clamp_min(irate(ceph_rbd_write_latency_count[30s]), 1) +\n irate(ceph_rbd_read_latency_sum[30s]) / clamp_min(irate(ceph_rbd_read_latency_count[30s]), 1)\n ) by (pool, image, namespace)\n)\n",
"format": "table",
"instant": true,
"intervalFactor": 1,
projects / ceph-ci.git / commitdiff