/tmp says 100%, du says different?

[myusername]

Anyone know if there is some sort of cache on the /tmp quota that anyone is aware of?

I have WHM showing 100% full for /tmp where

cd /tmp
du -sh *

shows only:

root@mybox [/tmp]# du -sh *

16K     lost+found
0       mysql.sock
4.0K    nobody-session-0.536175091591762
0       sess_14ea8906d6d1e102d993cff641f9013a
0       sess_1eb4dc7e3876974c0c99853ed96747ca
8.0K    spamd-15241-init
20K     spamd-15521-init

Any ideas why it still says full and how to get it to reflect an accurate % full?

[chirpy]

What does df give you?

[myusername]

Hi Chirpy,

Thanks for the suggestion Chirpy. I had to reboot it as a quick fix while I look for a solution. That cleared it up for the time being.

df currently lists the same thing that WHM does:

/dev/sda7 1035660 40732 942320 5% /tmp

Which leads me to believe before this incident it would have read

/dev/sda7 1035660 40732 942320 100% /tmp

My guess is there is a way to refresh the actual disk usage cache but I do not know what that command might be. There is a checkbox in tweak settings related to this but it did not solve the issue.

[myusername]

Well it is full again and here are the results requested:

This is the output before removing the files in /tmp:

root@someserver [~]# df

/dev/sda7              1035660    983052         0 100% /tmp

======
root@someserver [/]# du -sh *

928M    tmp

output after the files have been removed:

root@someserver [~]# df

Filesystem           1K-blocks      Used Available Use% Mounted on

/dev/sda7              1035660     35616    947436   4% /tmp

=======
root@someserver [/]# du -sh *

3.7M    tmp

Yet WHM still lists 100% and user sessions are void, and no temp files can be used, for example, uploading images via scripts until this cache gets cleared, which so far gets done by reboots. ;(

[chirpy]

The thing missing is what is filling up your /tmp partition that you have to keep clearing down? If you can fix that issue, other problems will probably go away?

[dgbaker]

Exactly what I was thinking. Next time also do an `ls -latr /tmp` to see what all is in there to know what to start looking at.

[myusername]

You guys think like me, fix the root of the problem. Problem is I did not program this server nor want to reverse engineer what has been done, and potentially break a highly customized cPanel server.

The developer will be back tomorrow, and we can get through the issue undoubtedly.

Still curious as to why removing the very large files that appear in /tmp do not immediately affect what WHM displays, (until reboot to get an accurate reading) and for that matter what web based users can do.

I can understand a web-based service check being cached and "displayed" to user or admin for performance issues, but when the "actual" content in /tmp is basically zero, preventing users from performing operations that require /tmp access is new to me.

Seems like they should be allowed /tmp access if the actual quota is near 1-5% regardless of the cached, web-based display quota of 100% says.

Thanks again.

[chirpy]

What you see as the seeming caching effect is actually chkservd not having run again after you cleared up after the problem. If it is chkservd, then you could restart it for an automatic refresh of the status page using:

/etc/init.d/chkservd restart

Then reload your WHM status page.

However, running out of space in your /tmp partition can cause significant problems and knock-on effects for the rest of the servers operating system and applications. It's not unknown for your server to hang, logins don't work, data corruption and a myriad of other problems, should any partition that is used runs out of space, and the /tmp partition can be doubly as important.

If it does happen again, I would suggest you do as dgbaker suggested and post the output from the ls command on this thread where people can help you understand what may be causing the initial problem.

[dgbaker]

Also note that in Tweak Settings that if you have the following unchecked, you can get very delayed information.

Disable Disk Quota display caching (whm will cache disk usage which may result in the display of disk quotas being up to 15 minutes behind the actual disk usage. Disabling this may result in a large performace degradation.)

[myusername]

I will do that the next time it happens.

As for the "Disable Disk Quota display caching" I tried that the first time for giggles, and it did not affect anything in WHM so far as displaying *current* accurate usage for /tmp. I will see if restarting the service check daemon does anything.

Thanks all.