[Pvfs2-users] the question about "pvfs2-server stop"
hxsrmeng at gmail.com
Thu Jan 17 18:13:25 EST 2008
Sorry if you get this post again. since I kept get response form
pvfs2-users-bouncer at beowulf-underground.org with the subject "the results of
your email commands" and my post sent back, and I didn't see my post in the
archives. So I guess something wrong with my email systems or email format?
I am trying to post it again.
When I am using
Stopping PVFS2 server: cat: /var/run/pvfs2.pid: No such file or
kill: usage: kill [-s sigspec | -n signum | -sigspec] pid | jobspec ...
or kill -l [sigspec]
I checked the mailing list archive and find the related posts attached
at the end of this post.
According to what Nathan Poznick said, in /etc/init.d/pvfs2-server
script, I commented "kill `cat $pidfile`" out, and added"killall
pvfs2-server", then ran
Stopping PVFS2 server: Terminated
Is this correct response?
Should it write some files out by "pvfs2-server stop"? How can I check
whether it does the right thing?
======================The relative posts==============================
Nathan Poznick wrote:
>Thus spake Tommy Butler:
>>Doesn't work :(
>> node1:~# /etc/init.d/pvfs2-server stop
>> Stopping PVFS2 server: cat: /var/run/pvfs2.pid: No such file or
>> kill: usage: kill [-s sigspec | -n signum | -sigspec] [pid | job]...
>> or kill -l [sigspec]
>Can you try running pvfs2-server inside strace to see if it's even
>attempting to write it out?
>Also, if the file cannot be written out for some reason, no error is
>thrown, so it's possible that's happening.
> * Manipulate the pid file. Don't bother returning an error in
> * the write stage, since there's nothing that can be done about it.
>static int create_pidfile(char *pidfile)
> return open(pidfile, (O_CREAT | O_WRONLY | O_TRUNC), 0644);
I don't really have time to debug. I rewrote the
/etc/init.d/pvfs2-server script to work around the problem. Instead of
creating a pid file and depending on the old "kill `cat $pidfile`"
convention, I simply issued a "killall pvfs2-server" call in the script
to stop the pvfs2-server process. Not exactly graceful I know, but
nonetheless effective; the only way this new way of killing off the
pvfs2-server process would be a problem is if someone were running
multiple pvfs2 servers-- which I see no reason for so doing.
Oh well, problem solved.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Pvfs2-users