#76 — hang/deadlock/livelock when reading from or writing to zfs filesytems.
| State | Rejected |
|---|---|
| Version: | 0.7.0 |
| Area | Functionality |
| Issue type | Bug |
| Severity | Medium |
| Submitted by | (anonymous) |
| Submitted on | Jul 15, 2010 |
| Responsible | Seth Heeren |
| Target release: |
—
|
Last modified on
Sep 19, 2010
by
Seth Heeren
I haven't quite narrowed down the cause(s) of the problem, but I can describe the symptoms. Once it happens, reads and writes simply stop happening. Some apps, like emacs, are effectively single threaded and block on the io, others aren't and keep working. Eventually X decides to do some io and then you're reaching for the reset button. Even running 'reboot' from a user shell will hang, probably because it's searching the path, which includes ~/bin which is on a zfs filesystem.
There's one case that I can reproduce at will. I record certain information every day in emacs planner-mode (what hours I worked on which projects, etc), and loading the daily planner file will invariably lock up emacs. I'll attach an strace of emacs, which is rather odd.
I started with the 0.6.0 rpm, then upgraded to 0.6.9 to test the latest rpm and in the hope that this would go away. Now I'm running the latest from the git repository with the same problem.
I've got a script that takes snapshots of the majority of my file systems at regular intervals, which I disabled temporarily to see if that helps, but the problem remains.
There's one case that I can reproduce at will. I record certain information every day in emacs planner-mode (what hours I worked on which projects, etc), and loading the daily planner file will invariably lock up emacs. I'll attach an strace of emacs, which is rather odd.
I started with the 0.6.0 rpm, then upgraded to 0.6.9 to test the latest rpm and in the hope that this would go away. Now I'm running the latest from the git repository with the same problem.
I've got a script that takes snapshots of the majority of my file systems at regular intervals, which I disabled temporarily to see if that helps, but the problem remains.
- Steps to reproduce:
- The simplest way that I have to reproduce this is to run M-x plan in emacs, which opens the day's planner file and does a few manipulations. Other things that have apparently set it off: offlineimap, rsync, mv, etc.
Added by
(anonymous)
on
Jul 15, 2010 08:07 AM
Here's a backtrace of zfs taken while emacs is hung: http://db48x.net/temp/zfs-fuse.gdb.log
Added by
Seth Heeren
on
Jul 15, 2010 09:46 AM
Responsible manager:
(UNASSIGNED) → sgheeren
For the record, now that we're back to the old fashioned juggling of repositories (why?) which version is that with?
Added by
Seth Heeren
on
Sep 19, 2010 06:34 PM
Issue state:
unconfirmed → rejected
closing because i recall that the issue was unrelated to zfs-fuse (see usergroup; i believe it was kernel + certain sata controllers)

