[CLI] Workaround to set a custom heap snapshot dir

[dokmic]

Summary

When Kibana is running inside a Docker container, the root filesystem is read-only. If we try to use --heapsnapshot-signal=SIGUSR2 in the NODE_OPTIONS and then send this signal to the Kibana process, the process crashes with an error complaining about the read-only filesystem.

The --diagnostic-dir parameter is supposed to set the destination folder for the heap snapshots, but there is a bug in Node <20 (nodejs/node#39493).

Since Kibana is running under non-root user, there is no way to work this around with mount --bind. And apparently, the Node 20 upgrade is not going to happen shortly (#162696). Due to those facts, the easiest option would be to use the writeHeapSnapshot() call.

The proposed workaround takes place only if both options --heapsnapshot-signal and --diagnostic-dir are present at the same time, and can be safely removed right in #162696.

Checklist

• This was checked for breaking API changes and was labeled appropriately

[elasticmachine]

Pinging @elastic/kibana-core (Team:Core)

[elasticmachine]

Pinging @elastic/kibana-operations (Team:Operations)

[jbudz]

Since Kibana is running under non-root user, there is no way to work this around with mount --bind. And apparently, the Node 20 upgrade is not going to happen shortly (#162696). Due to those facts, the easiest option would be to use the writeHeapSnapshot() call.

@watson are you planning on picking up the Node 20 PR in the near future? It looks mostly ready?

[watson]

@watson are you planning on picking up the Node 20 PR in the near future? It looks mostly ready?

I haven't had time due to serverless work. If someone wants to take it over I'll happily do hand over. Or I can probably find time this month.

[kibana-ci]

💛 Build succeeded, but was flaky

• Buildkite Build
• Commit: 2c26101

Failed CI Steps

• Serverless Osquery Cypress Tests #5
• Serverless Osquery Cypress Tests #6

Metrics [docs]

✅ unchanged

History

• 💛 Build #173888 was flaky 3f87c31

To update your PR or re-run it, just comment with:
@elasticmachine merge upstream