https://bugzilla.novell.com/show_bug.cgi?id=810239 https://bugzilla.novell.com/show_bug.cgi?id=810239#c0 Summary: midnight commander makes zfs-fuse userland not responding for approx. 30 min Classification: openSUSE Product: openSUSE 12.2 Version: Final Platform: x86-64 OS/Version: openSUSE 12.2 Status: NEW Severity: Normal Priority: P5 - None Component: Other AssignedTo: bnc-team-screening@forge.provo.novell.com ReportedBy: kaspernasebaer@gmx.de QAContact: qa-bugs@suse.de Found By: --- Blocker: --- Created an attachment (id=530394) --> (http://bugzilla.novell.com/attachment.cgi?id=530394) tail of strace log when mc hangs User-Agent: Opera/9.80 (Windows NT 6.1; WOW64) Presto/2.12.388 Version/12.14 When browsing huge datasets (>1TB) by using midnight commander, in some cases zfs-fuse userland is not responding anymore for some period of time (approx. 30min) and then recovers without any error messages in the logs. "# ls /pool" or "# zfs list" dont show any result and may be aborted by ctrl+C. Samba shares on the zfs volume are not accessible. Reproducible: Sometimes Steps to Reproduce: 1. create zfs-fuse pool containing a certain amount data and directories (~>1TB) 2. use midnight commander and navigate through directories 3. Sometimes the zfs-fuse filesystem stops responding. Actual Results: Midnight commander just "hangs" for a longer period of time and then continues without any error messages. The zfs-fuse driver remains active approx. 3.5% CPU. The hard disc drives do not appear to be busy during hang. Running midnight commander by strace shows that it appears to be calling "stat" and is just waiting for the result for approx. 30 min and then continues without showing any error. Expected Results: No hang or at least report an error/timeout. My impression is that this usually happens shortly after "chdir" is issued. I have added the tail of some strace output when running midnight commander. We are experiencing this bug on two different machines, both are 64bit and multi core and have 12-48GB of memory. One machine is using local discs, the other one is using an iSCSI device. The datasets are between 2TB and 48TB. No dedup and no compression is used. Updating to the most recent software from the openSuSE repos did not solve the problem. Shutting down and starting the zfs-fuse daemon of the 2TB dataset on local drives sometimes takes some period of time (1-2h) which appears quite long. Strangely, we were _not_ able to trigger this bug by running multiple processes in parallel like kernel builds or e.g. "find /pool -type d -exec ls -la '{}' \; >> /dev/null ; " or "find /pool -type f -exec stat '{}' \; >> /dev/null " or by using bonnie++. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.