vx_nospace error message

[Michael D. Zorn]

I'm running an application that reads a file and writes to another. All of a suddin, I'm getting this message:

msgcnt 6 vxfs: mesg 001: vx_nospace - /dev/root file system full (1 block extent)

/var/adm/syslog/syslog.log has baout the same thing:

(datestamp, hostname) vmunix: msgcnt 5 vxfs: mesg 001: vx_nospace - /dev/root file system full (1 block extent)

I don't understand "/dev/root file system".

#bdf
shows noting near full.

[A. Clay Stephenson]

Typically, this is a result of bringing the system up in lvm maintenance mode and then not rebooting. In any event, it means that your / filesystem is filling up. It might mean something as simple as /tmp isn't mounted so that temporary files are going into the /tmp directory as opposed to the /tmp filesystem.

[A. Clay Stephenson]

Oh, and eventhough no filesystems are full now doesn't mean that one wasn't earlier. This is very typical of temporary files. The files are open()'ed, unlink()'ed, and then written to and read from before finally being close()'ed --- or more like in this case, the process terminated. ONly then is the space returned to the filesystem. By the time you ran bdf, the crisis was over.

[Michael D. Zorn]

It's not maintenance mode - I haven't done that.

I did find a fairly large file: /tmp/wd_action.dbg, with a very recent timestamp (957k). But - I'm not debugging, and the program is not compiled with a debug option.

I'll see if I can see if /tmp is mounted OK - mnttab has /tmp on /dev/vg00/lvol4 ...

... so it should be mounted - everything else is.

The wd_action.dbg file has lots of lines like

----Monitor Start-----
--->setting action to (0)
----End Monitor Start
--->rw(): New action = <0>
--->rw():reread: action = <0>

.... but now I run the program again, and there's no error.

The input and output files are about the same size, ~230Mb.

How can I tell if /tmp is really mounted - outside of doing a umount / mount ?

[A. Clay Stephenson]

Just run the mount command w/o arguments. It will display all the mounted filesystems. My mention of /tmp was nothing more than a guess and should be taken as no more than that. Some thing was filling up the / filesystem and since that something also has returned the space that has all the markings of a temporary file. The / filesystem should be all be fixed in size once a system has been running a few days at most --- and that is since the OS load rather than the last reboot. It could also be that some configuration file or enviroment variable used by an application has changed so that files are not now written to the correct place. I would setup a cron job to run the date command and the bdf command perhaps every 5 minutes and append that output to a logfile and you may get a handle on what is happening. From reading your question, it's not clear if this application you are running is directly related to the problem or not. Your cron'ed bdf's could provide that data if you let it run for an hour or so before launching this application.

[tkc]

looks like your / directory is now mounted on /dev/root instead of /dev/vg00/lvol3. is that true? if it is true, you'll want to check that /etc/fstab correctly shows that /dev/vg00/lvol3 should
mount on / . If it does then:

#mv /etc/mnttab /etc/mnttab.old (so it can be recreated)

#mount -a (which will read /etc/fstab and rebuild /etc/mnttab)

Then check for correct mount later with bdf command.

[Michael D. Zorn]

Clay: 'mount' shows /tmp mounted. The machine has been running several weeks. The program giving the error (though it stopped doing that - no way to tell if it'll continue) runs for only 30 S. There's a utility that captures system statistics, like disk use, cpu load, &c, and you can set it to log every N seconds - I can't think of the name - do you know the one?

Just what is /dev/root (and /dev/rroot)? There are references on Google, and it seems to be the device file for root: ?

But 'df' shows no line for '/dev/root', as some of the examples I found on Google do. All I have are /stand, /var, /usr, /tmp, /opt, /home, and our local FS.

Is /dev a mountable FS in some systems?

tkc: I checked to see where / is mounted. It's OK. 'mount' shows / on /dev/vg00/lvol3, and 'll /dev/root' shows
brw-r----- 1 bin sys 255 root
(same for rroot)

[James R. Ferguson · Accepted solution]

Hi Michael:

> There's a utility that captures system statistics, like disk use, cpu load, &c, and you can set it to log every N seconds - I can't think of the name

# man sar

Regards!

..JRF...