Sometimes, logs are not shown

[dgageot]

To reproduce this, I run the micro-services in dev mode, interrupt it with ctrl-c and restart it in dev mode.

Sometime all the logs are there. Sometimes, it feels like skaffold is subscribing to the containers that are terminating and doesn't subscribe to those that starting.

[aeweidne]

I am also experiencing this with the Helm deployer, and I also had the guess it was listening to the terminating pods from the previous deploy.

[r2d4]

We are matching the containers by image name, but sometimes the image name is updated before the actual pod is re-rolled (i'm not sure how). We can fix this by directly referencing the image checksum. I tried to describe the issue here

#191

[aeweidne]

will #191 be reopened any time in the near future? having to trail the logs in a separate terminal ~50% of the time is one of the biggest things stopping skaffold dev from being a great experience for me

[dgageot]

@shimmerjs The thing is I don't think #191 would fix your issue. Can you describe your issue? Do you lose all the logs or only the first second?

[dgageot]

@shimmerjs Could you test #513 and see if it gets better?

[aeweidne]

@dgageot what I'm seeing is subsequent iterations of skaffold dev (whether interrupted by me or triggered by changes in workspaces) don't get any logs streamed about 50% of the time. The error rate seems to be tied to how fast K8s is tearing down the pods (i.e. this problem is lessened on my skaffold projects where the build and deploy time is longer) and occasionally I will get streamed the logs from the last few seconds of the terminating pod, which led me to this issue and thinking it was caused by the same root problem.

I will pull #513 and test it later today and get back to you, thank you!