Bug #4134
closed
mds: request locking hang under snaptests
Added by Greg Farnum about 11 years ago.
Updated almost 8 years ago.
Description
Haven't diagnosed this yet, but I ran into it while testing for #3819. Logs in kai:~/logs/4134.
First step, just figure out what the issue is. Then resolve or prioritize.
- Description updated (diff)
For reference, the hung request is client.4104:75.
- Status changed from New to In Progress
The request is attempting to read a snapshotted inode, but the inode is still in the LOCK_SNAP_SYNC state from when it was created (as part of cow_inode). I see a client_cap_release coming in that should apply to that inode, but it's being dismissed(?I think?) due to the wrong issue_seq. Checking out that angle now.
Oh, and actually the client is the one with the seq of 2 while the MDS only thinks it should have a seq of 1. Curiouser and curiouser...
- Priority changed from High to Normal
De-prioritizing this for now based on our meetings and discussion yesterday.
- Assignee deleted (
Greg Farnum)
- Status changed from In Progress to 12
- Priority changed from Normal to Low
- Status changed from 12 to Resolved
by commit 6cb9b01fb68eeb1449af0d145f31ee813a5316d2
Also available in: Atom
PDF