Skip to contentSkip to conversation

Can we increase the ping timeout from 20?

DISCUSSION Peter Senna Tschudin 1 participant 1 message Feb 5, 2025

lore
Share & tools
lore
#0 lore raw

Hi,

This is about:

commit ddfde25f16ba31fb480d2e83b29631aaa56526cb

Author: Petri Latvala <[email protected]>
Date:   Mon Sep 9 14:38:07 2019 +0300

runner: Add support for aborting on network failure

https://gitlab.freedesktop.org/drm/igt-gpu-tools/-/blob/master/runner/executor.c?ref_type=heads#L225

In that commit there was a seemingly arbitrary timeout of 20 seconds for the network to come to life after a resume. We are facing issues from time to time where it will take more than 20 seconds for the network to resume after a resume.

20 seconds seems to be adequate for many scenarios, but not all of them. I have experienced DUTS taking way more than that to get the network up during boot for reasons completely unrelated to what we are checking here.

So my question is what would be the upper reasonable limit for the timeout? Would bumping it to 60 seconds defy the original intention? Reading the commit message gives me the impression that as long as CI does not pull the plug because the network is down we are fine increasing the time out.

In other words, the upper limit seems to be the amount of time that CI waits for the network to come back to life after a suspend / resume.

Should I make this number configurable so that CI can match it with their timeout for the network to come up after a suspend/resume cycle? If not, can I bump it to say, 60 seconds?

Thank you!

Peter