You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A guess.
During the execution of inode.Parent.mu.Lock() and inode.Parent.mu.UnLock()
have other operations were running in parallel that caused modifications to inode.parent, make inode.parent to nil
wph95
linked a pull request
Sep 7, 2020
that will
close
this issue
we run the goofys in a docker container on the k8s.
we met this problem in our customer and our test cluster. not frequently (1-3 times a week)
Our Architecture
A container:
still create a folder to root with about 10 files every 30 seconds - 1 minute
B container: [have this runtime crash problem ]
just read the file A create. every 15-minute list root path to known which folder has been deleted
S3 bucket:
set a retention rule [7 day] on s3
guess when list root path, a lot of folders has been deleted. create a lot of ForgetInode op, make something race condition.
still working on finding a way to easily reproduce this problem and try to get the fuse_debug log when met this problem
I'm continuing to try to reproduce the bug reliably.
The text was updated successfully, but these errors were encountered: