Allow for custom dashboards

Signed-off-by: Boris Ranto <branto@redhat.com>

The node_exporter role should go before grafana

Signed-off-by: Boris Ranto <branto@redhat.com>

Rename the dashboards

Signed-off-by: Boris Ranto <branto@redhat.com>

ansible: Fix playbook.yml

Signed-off-by: Boris Ranto <branto@redhat.com>

Add srpm script

Signed-off-by: Boris Ranto <branto@redhat.com>

rpm: Remove unused files

Signed-off-by: Boris Ranto <branto@redhat.com>

Strip out graphite/collectd

Signed-off-by: Boris Ranto <branto@redhat.com>

Initial cleanup commit

- file hierarchy cleanup
- remove unnecessary files
- move links to their proper places in ansible directory

Signed-off-by: Boris Ranto <branto@redhat.com>

Merge pull request #209 from ceph/wip-branto

Downstream fixes

osd-node-detail: Fix value repetition

We did not sum the values for RAM usage so we ended up with a couple of
entries being shown for each RAM usage query.

Signed-off-by: Boris Ranto <branto@redhat.com>

ansible: Fix service_name indentation

The service_name for node_exporter was not indented properly.

Signed-off-by: Boris Ranto <branto@redhat.com>

Merge pull request #203 from ceph/wip-rpm-patch

rpm: use_epel is no longer defined

Merge pull request #208 from ceph/fix-throughput-units

Fix units used for throughput

Reviewed-by: Boris Ranto <branto@redhat.com>

Fix units used on throughput charts

Applies the same change to the prometheus based
dashboards as commit 239afda4debc80544112fe98d31af3f692678f57 did for
the graphite based dashboards

Signed-off-by: Paul Cuzner <pcuzner@redhat.com>

Merge pull request #204 from ceph/osd-info-updates

Multiple fixes to OSD information dashboard

Fix units used for throughput

Throughput units were using binary representation,
whereas the expected unit is decimal (i.e.
MB/s not MiB/s)

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1496186
Signed-off-by: Paul Cuzner <pcuzner@redhat.com>

Multiple fixes to OSD information dashboard

Bluestore tables and charts updated, including;
- switched units from ms to secs which shows us too
- changed metric from commit to KV latency
- updated thresholds in bluestore tables
- switched from rate to irate for bluestore metrics
- updated bluestore text box description

Signed-off-by: Paul Cuzner <pcuzner@redhat.com>

rpm: use_epel is no longer defined

Signed-off-by: Boris Ranto <branto@redhat.com>

Merge pull request #201 from zmc/wip-uneditable-dbs

Make dashboards uneditable by default

Merge pull request #200 from zmc/wip-osd-info-db

ceph-osd-information: Fix numerous bugs

ceph-osd-information: Fix numerous bugs

Too many to list here; almost every panel was
broken in some regard.

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #197 from ceph/add-rgw-latencies

Added RGW GET/PUT Latencies

Merge pull request #196 from ceph/iscsi-db-updates

Updated to use OS metrics from default scrape job

Merge pull request #199 from zmc/wip-cluster-db

ceph-cluster: Fix column styles for version tables

Merge pull request #198 from zmc/wip-iscsi-prom

Scrape iSCSI-related exporters

Make dashboards uneditable by default

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-prometheus: Scrape iscsi gateway exporter

Signed-off-by: Zack Cerza <zack@redhat.com>

playbook: Install node_exporter on iscsi gateways

Signed-off-by: Zack Cerza <zack@redhat.com>

Added RGW GET/PUT Latencies

Added multiple charts showing GET/PUT latencies
at overview and RGW detail levels. In addition
the failed HTTP request panel has been changed
from a singlestat to a graph to visualize the
failure rates across all RGW instances.

Signed-off-by: Paul Cuzner <pcuzner@redhat.com>

Updated to use OS metrics from default scrape job

All node_exporter scrapes are now done under the
same job (called node) so the dashboard now uses
an updated template query to identify the correct
host to pull out the OS metrics by iscsi gateway

Signed-off-by: Paul Cuzner <pcuzner@redhat.com>

ceph-cluster: Fix column styles for version tables

We were using 'id', not 'ceph_daemon' and as a result that column wasn't
showing up.

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #195 from zmc/wip-dashboard-fixes

Fix mon_server queries in network-usage-by-node and set home db

Merge pull request #187 from ceph/wip-rpm

rpm: Update spec file for recent changes

rpm: Modify node_exporter service name

Signed-off-by: Boris Ranto <branto@redhat.com>

ceph-grafana: Set the admin users's home db

If a human (or API request) changes the home dashboard for the admin
account, our method of setting it at the org level will no longer be
effective. Let's keep the admin user's home dashboard set to
ceph-at-a-glance.

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-node-exporter: Fix inaccurate task name

Signed-off-by: Zack Cerza <zack@redhat.com>

network-usage-by-node: Fix mon_server queries

The variable queries needed to drop 'mon.' from the mon names, and the
panel queries needed '[[mon_servers]]' to be wrapped in parentheses.

Signed-off-by: Zack Cerza <zack@redhat.com>

rpm: Modify container name/version

Signed-off-by: Boris Ranto <branto@redhat.com>

rpm: Update spec file for recent changes

Signed-off-by: Boris Ranto <branto@redhat.com>

Merge pull request #191 from zmc/wip-osp-fixes

Fixes for OSP

Reviewed-by: Boris Ranto <branto@redhat.com>

Merge pull request #192 from zmc/unpin-testinfra

Unpin testinfra

Reviewed-by: Boris Ranto <branto@redhat.com>

Unpin testinfra

1.14.0 has been released with the fix we needed.

Signed-off-by: Zack Cerza <zack@redhat.com>

node_exporter: Allow custom service name

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-mgr: Cope with differently-named containers

We were expecting ceph-mgr@hostname, but let's also look for
ceph-mgr-hostname.

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #188 from zmc/wip-containers

containers: grafana version; systemd unit comments

Merge pull request #190 from zmc/wip-db-fixes

Add some dashboard tests, and make them pass

dashboards/network-usage-by-node: Fix mon_hosts

Signed-off-by: Zack Cerza <zack@redhat.com>

dashboards: Remove cruft

Specifically, references to collectd, influxdb,
and hostnames.

Signed-off-by: Zack Cerza <zack@redhat.com>

dashboards: Drop 'DS_LOCAL' templating

Signed-off-by: Zack Cerza <zack@redhat.com>

Add some dashboard tests

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #179 from zmc/wip-drop-use_epel

Drop use_epel flag

Reviewed-by: Boris Ranto <branto@redhat.com>

Warn against modifying container systemd units

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-grafana: Default to 'latest' container version

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #183 from ceph/dashboard-refresh

Dashboard refresh

Dashboard uid metadata field removed

The presence of uid in the dashboard definition is causing
compatibility issues with moving to Grafana v5.1. By removing the
uid entry, the dashboards still work and can be migrated to 5.1

include pg unknown in pg state table

Include pg_unknown in PG status pie chart

Add RAM usage to OSD Node detail dashboard

Changes to health history and inclusion of pg information

Added pg distribution and pg per osd tables

add alert for high pgnums on an OSD

Update prometheus source name (align with ansible deployment)

capacity units updated for consistency

Drop encrypted osd panel, add dev-class piechart

Merge pull request #181 from zmc/wip-177

Work around #177

Reviewed-by: Boris Ranto <branto@redhat.com>

Merge pull request #185 from zmc/wip-grafana-fixes

Support Grafana 5.1.x

Reviewed-by: Boris Ranto <branto@redhat.com>

ceph-grafana: Restart container when creating

This is necessary so we can move to newer versions
 of the container image.

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-grafana: Use the correct UID for 5.1.x

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-grafana: Bump container version

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #167 from zmc/wip-mon-db-host

Monitor the dashboard host

Merge pull request #182 from zmc/wip-bluestore-selinux

ceph-collectd: Only set SEL context on filestore

ceph-collectd: Set SEL context on bluestore

... as well as filestore.

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #175 from zmc/wip-classic-fixes

Fixes for classic mode

Merge pull request #180 from zmc/fix-docker-tests

ceph-docker: Update tests for 'containerized' flag

Merge pull request #173 from zmc/wip-cpu-quota

ansible: Set CPU/RAM quotas on containers

ceph-docker: Update tests for 'containerized' flag

Signed-off-by: Zack Cerza <zack@redhat.com>

Work around #177

Drop use_epel flag

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #166 from zmc/wip-prom-pkg

Allow installing prometheus and node_exporter from packages

Reviewed-by: Boris Ranto <branto@redhat.com>

Merge pull request #176 from zmc/wip-fix-notification

ceph-grafana: Use 'admin_user', not 'user'

Reviewed-by: Boris Ranto <branto@redhat.com>

ceph-grafana: Use 'admin_user', not 'user'

We missed this when merging #165 & #170

Signed-off-by: Zack Cerza <zack@redhat.com>

tox.init: Unpin ansible for ansible-syntax

Signed-off-by: Zack Cerza <zack@redhat.com>

node_exporter: Fix RPM name

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-prometheus: Allow installation via packages

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-docker: Fail if using classic mode

Signed-off-by: Zack Cerza <zack@redhat.com>

classic: Fix pie charts w/ grafana 5.0

When legends are off, we're seeing "Invalid dimensions for plot, width =
179, height = 0"

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #170 from zmc/wip-notification

ceph-grafana: Add notification channel

Reviewed-by: Boris Ranto <branto@redhat.com>

ansible: Set RAM quotas on containers

We default to 4GB RAM, and we use double that value for swap.

Signed-off-by: Zack Cerza <zack@redhat.com>

ansible: Set CPU quotas on containers

We want to default to limiting each container to two cores. While docker
has a nicer way to do this, it isn't supported by ansible as of 2.5.2.
It's easy enough to use the slightly more awkward method, however.

Signed-off-by: Zack Cerza <zack@redhat.com>

playbook.yml: Add node_exporter tag

Signed-off-by: Zack Cerza <zack@redhat.com>

Merge pull request #165 from ceph/wip-branto

Add a couple of ansible fixes

ceph-grafana: Add notification channel

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-node-exporter: Open firewall ports

Signed-off-by: Zack Cerza <zack@redhat.com>

Add dashboard for monitoring the db host itself

Signed-off-by: Zack Cerza <zack@redhat.com>

ceph-prometheus: Add separate job for db host

So that we can monitor the db host's resources, without mixing that data
in with the ceph cluster's metrics.

Signed-off-by: Zack Cerza <zack@redhat.com>

playbook.yml: Deploy node_exporter on db host

Signed-off-by: Zack Cerza <zack@redhat.com>

Add a sample group_vars file

Signed-off-by: Boris Ranto <branto@redhat.com>

Fix grafana admin user/password setting

Signed-off-by: Boris Ranto <branto@redhat.com>

Extend purge.yml

Signed-off-by: Boris Ranto <branto@redhat.com>

Set home dashboard

Signed-off-by: Boris Ranto <branto@redhat.com>