[Bug 386080] New: dbench errors when used on NFS.
https://bugzilla.novell.com/show_bug.cgi?id=386080 Summary: dbench errors when used on NFS. Product: openSUSE 11.0 Version: Beta 1 Platform: x86-64 OS/Version: openSUSE 11.0 Status: NEW Severity: Major Priority: P5 - None Component: Kernel AssignedTo: bnc-team-screening@forge.provo.novell.com ReportedBy: pmorreale@novell.com QAContact: kernel-maintainers@forge.provo.novell.com Found By: Development dbench fails to properly complete its cleanup when used on NFS with openSUSE 11 Beta1. Repeat by configuring two Beta1 machines, a client and server and issue: % dbench -t 20 10 As dbench completes the run, it attempts (and fails) to remove files/directories. Rerunning the same against attached disk completes successfully. The issue can be seen on both Ethernet and IPoIB exports. Note that Trond just pushed a number of NFS patches to Linus last week, so these problems may be solved with that patchset. -- 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.
https://bugzilla.novell.com/show_bug.cgi?id=386080
User gregkh@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=386080#c1
Greg Kroah-Hartman
https://bugzilla.novell.com/show_bug.cgi?id=386080
User pmorreale@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=386080#c2
--- Comment #2 from Peter Morreale
What specific patches do you think solve this issue?
I don't know specifically, it was merely an observation based on watching the linux-nfs list. I did not review the patches. It smells like races in dcache handling, but I'm just not sure at the moment. I haven't had time to investigate at all. I just wanted to call attention to what I found. -- 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.
https://bugzilla.novell.com/show_bug.cgi?id=386080
User gregkh@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=386080#c3
Greg Kroah-Hartman
https://bugzilla.novell.com/show_bug.cgi?id=386080
User gregkh@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=386080#c4
Greg Kroah-Hartman
https://bugzilla.novell.com/show_bug.cgi?id=386080
User nfbrown@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=386080#c5
Neil Brown
https://bugzilla.novell.com/show_bug.cgi?id=386080
User pmorreale@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=386080#c6
Peter Morreale
Is this actually a regression? i.e. has dbench over NFS ever worked?
I seem to recall that it unlinks files that it still has open, thus leaving .nfsXXXX files lying around which causes 'rmdir' to fail.
So: has this ever worked?
Excellent point. I assumed that it did when in fact dbench also fails with a SLED server and SLES 10 SP1 client. THat is most likely the issue. I created a simple test case that creates a dir/file, unlinks, closes and rmdir's the directory. If I omit the close, then the test case fails on NFS. I suspect that dbench is not closing the open files prior to the rmdir(2)s. I'm setting the status to INVALID right now, I'm not sure whether this is correct as I do not know the proper protocol for bug flow. This is clearly not a NFS bug. Is there a website documenting our bug resolution flow? thx -- 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.
participants (1)
-
bugzilla_noreply@novell.com