gluster-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gluster-devel] Add me too for lockup of system for very simple Glus


From: Mark Mielke
Subject: Re: [Gluster-devel] Add me too for lockup of system for very simple GlusterFS config...
Date: Sun, 06 Sep 2009 04:44:59 -0400
User-agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.1) Gecko/20090814 Fedora/3.0-2.6.b3.fc11 Thunderbird/3.0b3

On 09/06/2009 03:42 AM, Anand Avati wrote:
On 9/6/09, Mark Mielke<address@hidden>  wrote:
I used a very simple configuration. The server has a posix volume with
a server volume. The client has single client brick with a
cluster/replicate. I'd show you the volume files, but the server won't shut
down and restart to let me show you them, and I'll have to go in to the
server room to restart it.

If you tell me what commands you would like me to run, I'll try to help out
in diagnosing the problem. I am new to GlusterFS, and this was really my
first test.
is it possible to get a dmesg output somehow? anything on the console?
is this the system running storage/posix (server)? or the fuse mount
(client) or both?

Not any more - I got impatient and did 'kill -9 -1' to see if that would help it shut down, but it only seems to have killed sshd. I will try it again tomorrow.

I was using glusterfsd with storage/posix and mount -t glusterfs on the same machine.

Another odd symptom for me - which I do not recall others having: I had /export/gluster-test as the storage/posix directory, I had mounted as a client on /tmp/t, and while cd'ed into /tmp/t, I could change files and see them without any hang. It only hung when I tried to cd to /export from a separate window to see what effect my changes in /tmp/t had. Weird, eh? After this, any SSH in to /home/markm worked, and su to root worked, but a cd to / or /export hung. Calls to df would show several lines of output and then hang (presumedly on the /export/gluster-test line?).

That cd / froze, when GlusterFS was exporting /export/gluster-test as /tmp/t, makes me doubt that GlusterFS is directly responsible, although it seems clear that it was GlusterFS that triggered the problem since I have had no problems with the system before I tried this test.

One possibly "weird" configuration I have is that /tmp is a tmpfs file system, but even with this configuration, I've never had problems mounting any other file system as /tmp/t, and this is a common pattern for me. For example, I often mount /tmp/t as the USB drive to rsync data to as a alternate backup strategy.

--
Mark Mielke<address@hidden>





reply via email to

[Prev in Thread] Current Thread [Next in Thread]