Github user joshelser commented on a diff in the pull request:
https://github.com/apache/accumulo/pull/220#discussion_r102778301
--- Diff:
server/master/src/main/java/org/apache/accumulo/master/metrics/Metrics2ReplicationMetrics.java
---
@@ -58,6 +59,7 @@ protected void snapshot() {
registry.add(PENDING_FILES, getNumFilesPendingReplication());
registry.add(NUM_PEERS, getNumConfiguredPeers());
registry.add(MAX_REPLICATION_THREADS, getMaxReplicationThreads());
+ registry.add(LATENCY, getLatencyInSeconds());
--- End diff --
You're collecting the "maximum" latency in the system which isn't ideal as
it paints a very skewed picture of the actual system.
For example, if you have a single file which cannot be replicated (say,
it's corrupted) it would have a very long latency. The rest of the files may be
replicating in a (hypothetically) 5seconds. The maximum latency would continue
to increase without bound, indicating to an operator that the system is not
healthy, when the system is actually 99% healthy.
Hadoop metrics2 exposes a histogram class for metrics which automatically
aggregates min, median, avg, max as well as percentiles (e.g. p75, p85, p95,
p98, p99).
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
