Fork me on GitHub
#clojure-uk
<
2020-02-01
>
dharrigan08:02:58

Good Morning!

👋 4
dharrigan19:02:01

Does anyone have any good recommendations for monitoring gc, how stressed the jvm is, potential early warning signs. I'm exposing a lot of metrics via prometheus and jmx.

3Jane19:02:35

grafana + prometheus is a classic duo

3Jane19:02:18

is that what you mean? Or are you looking for alerting thresholds and which metrics to observe?

dharrigan19:02:22

Using grafana + prometheus, just charting some metrics now, looks like my JVM is taking about 0.1ms to do a GC (using G1)

rickmoynihan09:02:26

Other candidate options: - datadog + their jmxagent - jvisualvm (free) can at times be a little awkward to connect to remote vm’s; but it’s possible… via an ssh tunnel/jmx combo etc. - jstat -gccause <PID> 1000 > /tmp/gclog.txt

dharrigan10:02:08

I'll look at datadog

dharrigan10:02:42

Having a gclog is a great idea, I think it can do truncation, or rollover and delete past entries. Probably.

Wes Hall20:02:29

@dharrigan Censum from jClarity is a nice tool for this, but it's commercial with 7 day trial. Oddly enough, it was one of the founders of this company that convinced me to start looking at clojure 🙂

dharrigan20:02:56

I looked at it earlier. It's very very expensive

dharrigan20:02:47

No way I could float that past the boss ATM :white_frowning_face:

Wes Hall20:02:38

I know the feeling re: keeping budgets tight. Might be some value in playing with the trial to see if you can get any quick insights.

Wes Hall20:02:10

Is jprobe still a thing?

Wes Hall20:02:01

I spent days with that way back when.

dharrigan20:02:59

I'll do some research and digging about. Thanks! 😁

Ben Hammond21:02:46

yourkit is still a thing

dharrigan06:02:19

Thanks! I forgot about YourKit!