[
https://issues.apache.org/jira/browse/CASSANDRA-20547?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dmitry Konstantinov updated CASSANDRA-20547:
--------------------------------------------
Description:
During a latency issues investigation it is useful to see when JVM is paused by
some reasons. There are JVM JMX metrics for GC but GC is not only the reason
for pauses, it can be due to too high CPU, noisy neighbour, VM movement, VM
over-provisioning, safepoints, etc.
https://github.com/giltene/jHiccup agent suggest a way to measure it. The agent
itself does not provide JMX metrics but we can repeat the same idea in our
code, it is a simple idea: run a thread which regularly go to sleep for a
configured time interval and then we compare the actual sleep time vs real
time:
https://github.com/giltene/jHiccup/blob/master/src/main/java/org/jhiccup/HiccupMeter.java#L484
We can report metrics like max pause per time internal and sum of pauses per
interval.
was:
During a latency issues investigation it is useful to see when JVM is paused by
some reasons. There are JVM JMX metrics for GC but GC is not only the reason
for pauses, it can be due to too high CPU, noisy neighbour, VM movement, VM
over-provisioning, safepoints, etc.
https://github.com/giltene/jHiccup agent suggest a way to measure it. The agent
itself does not provide JMX metrics but we can repeat the same idea in our
code, it is a simple idea: run a thread which regularly go to sleep for a
configured time interval and then we compare the actual sleep time vs real
time:
https://github.com/giltene/jHiccup/blob/master/src/main/java/org/jhiccup/HiccupMeter.java#L484
> Report Cassandra server pauses as metrics using jHiccup-like logic
> ------------------------------------------------------------------
>
> Key: CASSANDRA-20547
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20547
> Project: Apache Cassandra
> Issue Type: New Feature
> Components: Observability/Metrics
> Reporter: Dmitry Konstantinov
> Assignee: Dmitry Konstantinov
> Priority: Normal
>
> During a latency issues investigation it is useful to see when JVM is paused
> by some reasons. There are JVM JMX metrics for GC but GC is not only the
> reason for pauses, it can be due to too high CPU, noisy neighbour, VM
> movement, VM over-provisioning, safepoints, etc.
> https://github.com/giltene/jHiccup agent suggest a way to measure it. The
> agent itself does not provide JMX metrics but we can repeat the same idea in
> our code, it is a simple idea: run a thread which regularly go to sleep for a
> configured time interval and then we compare the actual sleep time vs real
> time:
> https://github.com/giltene/jHiccup/blob/master/src/main/java/org/jhiccup/HiccupMeter.java#L484
> We can report metrics like max pause per time internal and sum of pauses per
> interval.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
