mgr's perf-schema bridge silently drops labeled counters on the way
out, so shipping the per-bucket LC counters up through MgrReport just
costs ingest memory for data mgr can't expose anyway. ceph-exporter
already handles labeled counters via the daemon admin socket, so make
that the only path.
Signed-off-by: Matthew N. Heler <matthew.heler@hotmail.com>
Lifecycle metrics are stored as labeled performance counters in memory. All counters are lost when the Ceph Object Gateway restarts or crashes.
+Since ``ceph-mgr`` cannot expose labeled counters today; use the per-host ``ceph-exporter`` daemon to scrape these metrics.
+
Lifecycle Counter Cache Size & Eviction
----------------------------------------
const std::string rgw_lc_counters_key = "rgw_lc_per_bucket";
void add_lc_counters(PerfCountersBuilder *pcb) {
- pcb->set_prio_default(PerfCountersBuilder::PRIO_USEFUL);
+ /*
+ * mgr can't expose labeled counters via its prometheus module, so
+ * register at PRIO_DEBUGONLY to keep them out of MgrReport under the
+ * default mgr_stats_threshold.
+ */
+ pcb->set_prio_default(PerfCountersBuilder::PRIO_DEBUGONLY);
pcb->add_u64(l_rgw_lc_per_bucket_start_time, "start_time",
"LC processing start timestamp (Unix epoch seconds)");