xfs
[Top] [All Lists]

Re: rm hanging intermittently

To: "Eric Sandeen" <sandeen@xxxxxxx>
Subject: Re: rm hanging intermittently
From: "Richard Smith" <rgsmith72@xxxxxxxxxxxxx>
Date: Sat, 11 Jan 2003 22:38:51 -0800
Cc: <linux-xfs@xxxxxxxxxxx>
References: <Pine.LNX.4.44.0301112201060.27913-100000@stout.americas.sgi.com>
Reply-to: "Richard Smith" <rgsmith72@xxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
    I don't currently have the debugger built into this kernel, but I will
create a kernel with it installed and post the backtrace the next time the
process hangs. To answer your question, we have waited up to an hour for the
rm process to return. This system has two xfs partitions installed, a
smaller 36GB and a larger 450GB. It seems that when the rm process is hung
on one partition, write operations are also halted on the other partition.
Is XFS single threaded in this manner?
    Another data point is that 3 out of the last 4 times, the rm processes
were hung just after 5pm. I don't see any cron jobs running at this time of
day that would affect the filesystems, so this may be a coincidence.

Rick
----- Original Message -----
From: "Eric Sandeen" <sandeen@xxxxxxx>
To: "Richard Smith" <rgsmith72@xxxxxxxxxxxxx>
Cc: <linux-xfs@xxxxxxxxxxx>
Sent: Saturday, January 11, 2003 8:02 PM
Subject: Re: rm hanging intermittently


> Hi Rick -
>
> If you have kdb, it might be interesting to look at the backtrace of
> the stuck rm process, to see where it's at:
>
> kdb> btp <pid>
> kdb> go
>
> How long have you waited for it to get "unstuck?"
>
> -Eric
>
> On Sat, 11 Jan 2003, Richard Smith wrote:
>
> > Hello,
> >     I am experiencing a problem where a "rm -rf" command acting on a
450GB XFS partition will hang approximately once per day. The system is
running a daemon that issues rm commands throughout the day and 99% of
commands proceed without hanging. The system is using the 2.4.20-rc1 XFS
kernel compiled with gcc 2.96 on a dual P4 xeon server. The XFS partition
uses the linux software raid at level 0 with the XFS partition built on the
resultant device.
> >     Once the rm process is hung, other processes trying to access the
XFS filesystem are blocked, but the system is still responsive. Killing the
hung process frees up the filesystem and re-issuing the identical command
that originally hung will proceed without failure. Could this be a similar
problem as the one that caused hanging processes where the kernel was
compiled with gcc 2.95? Any help appreciated.
> >
> > Rick Smith
> >
> >
> > [[HTML alternate version deleted]]
> >
> >
>


<Prev in Thread] Current Thread [Next in Thread>