andygrove opened a new pull request, #4282:
URL: https://github.com/apache/datafusion-comet/pull/4282

   ## Which issue does this PR close?
   
   Supersedes #4276 (combines that change with the heap cap needed to actually 
pass on `ubuntu-latest`).
   
   Closes #.
   
   ## Rationale for this change
   
   The ASF runs-on.com app is currently disabled, so PR jobs that target the 
runs-on.com runner pool queue indefinitely and we cannot merge anything. #4276 
gates the runs-on.com selector behind `vars.USE_RUNS_ON` so jobs fall back to 
`ubuntu-latest` when the variable is unset. That is enough to dispatch jobs 
again, but several jobs then OOM on `ubuntu-latest` (16 GB) because the 
`verify-benchmark-results-tpch` and `verify-benchmark-results-tpcds` 
data-generation steps were calling `MAVEN_OPTS='-Xmx20g' ../mvnw exec:java 
...`. Asking for a 20 GB JVM heap on a 16 GB box guarantees an OOM.
   
   ## What changes are included in this PR?
   
   On top of #4276's `USE_RUNS_ON` gate:
   
   - Add a job-level `MAVEN_OPTS: -Xmx6g -XX:MaxMetaspaceSize=1g` env to 
`verify-benchmark-results-tpch` and `verify-benchmark-results-tpcds`. This 
applies to every `mvnw` invocation in those jobs (build, data gen, query test).
   - Drop the inline `MAVEN_OPTS='-Xmx20g'` prefix from the two TPC-H / TPC-DS 
data-generation steps so they pick up the job-level value. SF=1 data gen needs 
nowhere near 20 GB.
   
   The `linux-test` matrix already pins `MAVEN_OPTS="-Xmx4G -Xms2G ..."` inside 
the `java-test` composite action, so it is not changed here.
   
   ## How are these changes tested?
   
   CI on this PR runs the same affected jobs against `ubuntu-latest` (since 
`USE_RUNS_ON` is unset on this branch / on forks), exercising the new heap cap 
end-to-end.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to