Need to investigate segment deletion

[rbouqueau]

Imported from the old issue tracker. Created by Jack:

I see a clear bump in latency once segment deletion starts.

That is eventually partially caught up, but it takes a fairly long time.

This session was run with segmentLife=20000 and I’ve seen the same bump with other values.

For MMSYS we will probably just set segmentLife to be greater than our session duration. We produce about 5 MB/s, so there shouldn’t be a problem keeping tat in memory for a few minutes.

[rbouqueau]

Today's call: Jack to test again if the issue is still present.

[jackjansen]

Pushing this back to @rbouqueau . Romain: can you add a --timeshift_buffer_ms argument to testlatency.py, and then test with a reasonably low value, maybe something like 10 seconds, and a session duration of 60 seconds?

Then, if the issue still exists you should start seeing extra latency at t=20s, t=30s, etc.

[rbouqueau]

I coded the option locally but I run into the Inconsistent API version between compilation (539297572) and runtime (539297570) option. Is the cwipc version up to date API-wise?

[rbouqueau]

Jack is 99% that the problem has gone away. I need to QA that though.

[rbouqueau]

Jack to QA the PR #87 .

Romain to test by reading the relay logs that segments are deleted.

[rbouqueau]

Now I can run. Is ok.

[jackjansen]

Ran a couple of sessions in repo 2024-spirit-lldash. This issue is confirmed fixed: there is no extra latency incurred when deleting segments.