Could not pop into the server

[viraj]

I use a server that has 147 GB OF Hard disk and 4 GB Of memory… I have around 500 users configured in the server, yesterday I came a across a funny problem. In the statics in the server showed that the hard disk was full. I did a DF –M in the server and it shows that only 50% of the partition was full where Zimbra has been installed.
Problems that came up was
1. Could not pop into the server
2. When email were sent to any used it never go delivered to the user.
3. The mail that was sent to the server during this time has gone missing.
What I did was deleted some email and freed up some space on the server. And restarted the server then it came right…
Done anyone have any idea what is the problem. …

[ArcaneMagus]

What is the output of the following two commands?

mount -t ext3
df -mlP

Or if you are running on OS-X:

mount -t hfs
df -ml

The zmdisklog uses these commands to generate the statistics.

[viraj]

Just for information i did not still upgrade ZCS release 5.0.10.. i am still using 5.0.6

This is what i got for :mount -t ext3

root@mail:~# mount -t ext3
/dev/cciss/c0d0p2 on / type ext3 (rw,errors=remount-ro)


For : df -mlP ( the out put is given below... but now i have freeup space by deleting email from the server )

Filesystem         1048576-blocks      Used Available Capacity Mounted on
/dev/cciss/c0d0p2       134024     42655     84561      34% /
varrun                    2026         1      2026       1% /var/run
varlock                   2026         0      2026       0% /var/lock
udev                      2026         1      2026       1% /dev
devshm                    2026         0      2026       0% /dev/shm

[ArcaneMagus]

Well as you said the disk is only at 34% now...so I have no idea how the logger service was showing a full disk. For each line returned by the mount command it parses the output of the df command to get how much of it is used. If you are interested in the specifics the file that does this is /opt/zimbra/libexec/zmdisklogger

As far as your issue is concerned it probably relates far more to this mount option: "errors=remount-ro"
This means that when I/O errors are detected the file system is remounted as a read only filesystem...which would certainly give the symptoms you described.
/dev/cciss/c0d0p2 from what I am reading is generally a compaq raid controller (see: Linux Kernel Documentation :: cciss.txt)

I would try checking this filesystem for faults as it might be failing. At the very least make sure your backup system is working and tested...