osd/perf_counters: raise prio of before queue op perfcounter

We have no a metrics about osd msgr in prometheus metrics,
osd_op_before_queue_op_lat can be as this metrics, it represents
latency of op from recv_stamp to before queue. This section of
work is executed by the msgr-worker thread.

Signed-off-by: Yite Gu <yitegu0@gmail.com>

diff --git a/src/osd/osd_perf_counters.cc b/src/osd/osd_perf_counters.cc
index 588808d6467e1ba2dd67259509af185f730a7b5f..3e14244329eb1acfefd94338da2472da2de97f7f 100644 (file)
--- a/src/osd/osd_perf_counters.cc
+++ b/src/osd/osd_perf_counters.cc
@@ -123,13 +123,13 @@ PerfCounters *build_osd_logger(CephContext *cct) {
   osd_plb.add_time_avg(
     l_osd_op_rw_prepare_lat, "op_rw_prepare_latency",
     "Latency of read-modify-write operations (excluding queue time and wait for finished)");
+  osd_plb.add_time_avg(l_osd_op_before_queue_op_lat, "op_before_queue_op_lat",
+    "Latency of IO before calling queue(before really queue into ShardedOpWq)"); // client io before queue op_wq latency
 
   // Now we move on to some more obscure stats, revert to assuming things
   // are low priority unless otherwise specified.
   osd_plb.set_prio_default(PerfCountersBuilder::PRIO_DEBUGONLY);
 
-  osd_plb.add_time_avg(l_osd_op_before_queue_op_lat, "op_before_queue_op_lat",
-    "Latency of IO before calling queue(before really queue into ShardedOpWq)"); // client io before queue op_wq latency
   osd_plb.add_time_avg(l_osd_op_before_dequeue_op_lat, "op_before_dequeue_op_lat",
     "Latency of IO before calling dequeue_op(already dequeued and get PG lock)"); // client io before dequeue_op latency