lighthouse: Fatal trace logic error - expected start event, got X
I attached the assets from a LR run that resulted in this error message.
I’d provided the lhr too, but it’s on a remote machine, too big and scp
keeps crashing (side note, what???)
Some notes:
- could not get this to happen locally, neither via
node lighthouse
nor a local WRS. Only seems to happen in production LR environments - provided assets was from running against bloomberg.com. @exterkamp has a whole list of of other URLs that it happens on too.
- does not happen in prod LR
- happens on ~5% of requests in canary LR
Related issues
About this issue
- Original URL
- State: closed
- Created 5 years ago
- Reactions: 9
- Comments: 27 (3 by maintainers)
Hi, I am also getting this error when I run the Lighthouse test from the CI. It never happens when I run the test locally. This error appears in the Diagnostics of performance metrics for:
Is there a work-around I can put in place?
We expect this has been fixed by #9785 and #9786. The relative silence since that release went out is reassuring too, so we’ll go ahead and close this out but if anyone runs into this in >=6.0.0 and can provide a trace, we’d love to hear about it 😃
Kinda, but if it’s definitely a fatal trace logic error for trace events to overlap like that. Later logic will generate nonsensical numbers if there are two events executing at once like this or the events aren’t directly nested within their parents. We would have to force the
X
event to end at the timestamp of theB
event if we go this route.We should definitely add a test for this case too so something would’ve caught it I can get on that 😃
@christian-bromann not that I’m aware of. You’ve already filed in the correct place though for it to be looked into.
Thanks @benschwarz there was a lot more discussion around these cases in https://github.com/aslushnikov/tracium/issues/8 which fixes will be making their way over for our tracehouse refactor in the next few weeks.