-
Notifications
You must be signed in to change notification settings - Fork 4.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Error: image cache/mysql:pullthrough not found #8399
Comments
@miminar any idea why docker registry is falling to v1 registry, this is what I see in the logs:
It's downloading the image just fine, but suddenly it reaches for v1 registry, any ideas? |
I have the full docker registry log, if needed. |
If this is a flake that will affect users in production it's p0 (until we prove otherwise). |
After several investigations the problem I've managed to identify is this error:
The #8461 wrapped image-import and docker pull commands into |
This looks like a different error:
Looking into it. |
Here's another: https://ci.openshift.redhat.com/jenkins/job/merge_pull_requests_origin/5597/console (PR #8438) |
@miminar good catch, now that we have the import output this is what we got:
To nail the problem we'd have to log entire traffic between us and docker registry. It looks like docker registry is sometimes blocking us, requiring auth (?), because I can't seem to find any other reasonable explanation. Especially that the image we're trying to get is public, even more docker-provided. @smarterclayton @pweil- @miminar thoughts? |
For no 1. we could modify our tests in such a way that if it detects unauthorized error it'll retry, no idea how to address no 2. without code modifications, yet. |
Has the integration test flaked recently?
|
Can we get more tracing in those cases? Possibly the entire response body On Apr 18, 2016, at 5:23 AM, Maciej Szulik notifications@github.com wrote: For no 1. we could modify our tests in such a way that if it detects — |
That's what I was thinking about, but not sure if it's possible post-mortem. I'll dig into it. |
This may be related - saw this in against our registry
Which correlates to
|
This kind of looks like a race condition. |
I think this is fixed in #8558, the docker hub occasionally returns 401 on import. Need to clean the fix up and make it even more targeted (discuss whether to preserve this for DTR, which I bet has this problem as well). |
I'm reopening for the time of debugging the failure in https://ci.openshift.redhat.com/jenkins/job/test_pull_requests_origin_integration/301/consoleFull |
@smarterclayton after careful investigation the only two operations we do not retry are store.Open and io.CopyN from
The question is, should we retry, since apparently this is hitting the issue the great @liggitt identified, about hitting a clock-skewed server. |
What backs the input stream that is returned by open? On Apr 27, 2016, at 6:14 AM, Maciej Szulik notifications@github.com wrote: @smarterclayton https://github.com/smarterclayton after careful
and io.CopyN
from pullthroughBlobStore. The logs evidently shows the io.CopyN is the culprit in the output you've pointed me to, see: time="2016-04-25T15:36:57.066474262Z" level=info msg="Copying 217 The question is, should we retry, since apparently this is hitting the — |
Per this comment I'm proposing to wait a bit and see what happens and how fast their fix will be actually live. Additionally talking to @smarterclayton on IRC he's not too much comfortable with us retrying reading bytes. |
The same problem
appeared in https://ci.openshift.redhat.com/jenkins/job/test_pull_requests_origin_integration/450/ (the linked PR). |
I think this is resolved. |
I wouldn't be that sure, esp. that the issue is silent for past month... I'd love to be that way... |
I added retries to a few other places - I think this is fixed upstream, because I haven't seen retries in the logs when I've been looking at other issues. |
…that'll also fix travis
Potential flake. Can be seen here: https://ci.openshift.redhat.com/jenkins/job/merge_pull_requests_origin/5534/console
The text was updated successfully, but these errors were encountered: