Fix Multiple GC declared error when running Druid Cluster#17078
Fix Multiple GC declared error when running Druid Cluster#17078kgyrtkirk merged 3 commits intoapache:masterfrom
Conversation
|
I think the default configs should not set such a thing - unless its really beneficial. Its mentioned in https://druid.apache.org/docs/latest/operations/basic-cluster-tuning/ ; it seems like it was added as a suggestion in this PR. I looked at
I agree that the GC algo selection should be removed - as its important to ship something which works out of the box with the least amount of issues. |
|
would you like to remove |
|
@kgyrtkirk Thanks for the quick reply. I have removed the declaration in all files under |
abf623d to
f2b4250
Compare
|
note: there is no need to force push the branch as merge will be done with a squash (don't rebase your downstream branch) |
| -XX:+UseG1GC | ||
| -XX:MaxDirectMemorySize=128m | ||
| -XX:+ExitOnOutOfMemoryError | ||
| -XX:+UseG1GC |
|
hi @kgyrtkirk Thanks for approving my PR, can I trouble you to merge this PR now that the CI has passed? 😄 |
|
Thank you @GWphua for fixing this! |
Description
This problem surfaced when I am attempting to get MM-less Druid running on ZGC with Kubernetes.
After declaring
-XX:+UseZGCunderJAVA_OPTSfor my services, I faced the following error for only the Coordinator, Overlord and Router:"Error occurred during initialization of VM, Multiple garbage collectors selected".
It turns out that the declaration of the G1GC in the cluster configuration causes the error to surface. I also find it weird that this G1GC configuration is only in the Coordinator/Overlord and Router configuration. By removing the G1GC configuration, I am able to run my cluster with ZGC.
Since G1GC is the default option, there will be no change to Druid's default GC with this removal.
Fixed the multiple GC declared error.
Release note
Allow Druid to run non-G1 Garbage Collectors with
JAVA_OPTS.This PR has: