Join GitHub today
GitHub is home to over 36 million developers working together to host and review code, manage projects, and build software together.Sign up
Don't try to get the file name of purged files #4066
This has been bugging me for a while, but I only just now had a chance to track down the causes.
In my local instance, I was able to find a few places where this can occur. The simplest is to modify
@bgruening I think @dannon is right. The tracebacks in #1765 are due to calling
May 15, 2017
5 checks passed
Sorry, I was away at dinner (you'd be amazed how much sushi a 4 year old can eat).
@nsoranzo: There are oddly no messages in the logs, even at the debug log level. That's one of the more annoying things, since otherwise I'd at least see it when it happens rather than them needing to report it to me.
@dannon: Regarding #1765, I see that all the time and it's not directly related to this since 99% of the time the datasets are not in fact deleted or purged, the NFS is just slow and they haven't been flushed to disk by the cluster nodes yet. But yeah, trying to get a purged object's file name will cause the same sort of problem. I wonder, though, if it's really correct to fix this in the object store. I would think that's just hiding the fact that upstream functions are asking for file names when they shouldn't.