Hi there,

Recently installed Centos 5.6 in order to install Plesk on. Not familiar with Linux partitioning so left the default values at install with a few GB of unpartitioned space left free. Followed the instructions on the Plesk installation manual for adding a seperate partition for /tmp - http://kb.parallels.com/en/1410

Then later on I couldn't get in to the server locally via the GUI (which was already switched on), it just had a black screen so I powered the machine off. It then wouldn't boot back up. When it got to e2fsck I got a [Failed]. I also got the same message as in this thread.

I'm now at a remote location. When logging in as root and trying to run fsck to find and fix any disk problems a remote engineer has found that it won't unmount the partitions so I can't really run fsck on it. He's forced the machine to boot for me and I have SSH access but if I reboot the machine I'm quite sure I won't be able to get in again from here.

Is there something wrong with my fstab maybe?

/dev/VolGroup00/LogVol00 / ext3 defaults 1 1
LABEL=/boot /boot ext3 defaults 1 2
tmpfs /dev/shm tmpfs defaults 0 0
devpts /dev/pts devpts gid=5,mode=620 0 0
sysfs /sys sysfs defaults 0 0
proc /proc proc defaults 0 0
/dev/VolGroup00/LogVol01 swap swap defaults 0 0
/filesystems/tmp_fs /tmp ext3 noexec,nosuid,loop 1 1

Few things really

There's absolutely nothing keeping you from familiarizing yourself with what Linux system administration encompasses. Centos is the Red Hat Enterprise Linux derivative and RHEL comes with extensive documentation. Reading documentation may seem like a drag but the investment is worth it in terms of building up knowledge and server reliability, integrity and safety. If you don't have a server to play with at least have a staging server, workstation or virtualization guest to test things out in. Best practices state servers come with the least amount of packages installed necessary to fulfill their tasks. X server and Desktop Environments you run in runlevel 5 generally don't belong on a server. Also search LQ for partitioning threads because opinions differ. Generally speaking a "/ /home swap" layout is not as robust and configurable as the "/boot / /usr /tmp /var" one. In short separating the main file system directories allows you not only to use different mount flags but also keeps the system from keeling over when say /var gets filled. LVM makes sense if you 0) use dm-crypt or ecrypt-like encryption or 1) must be able to re-allocate storage on-the-fly, which doesn't make sense for say "/" if /usr, /home, /tmp and /var are separate partitions.


Basically it offers a file system-in-a-file loop/bin alternative for the time being but that doesn't trump actually having a separate /tmp partition. Unless you've past the point of no return if you agree separate partitions are better you could nuke the installation and start over?


Best not boot into a GUI after install if Xorg needs to be configured still. /etc/inittab holds the default runlevel to boot into: change from 5 to 3 or else change /boot/grub/grub.conf and add " 3" at the end of the kernel line.


Ext3 often is rugged enough to survive a file system getting umounted improperly but powering off properly also ensures the file system dirty flag gets cleared so you don't need fsck on reboot. And what you think is an error message is just instructions asking you to run 'fsck'. To force checking all filesystems at (re)boot you could add " forcefsck" at the end of the kernel line. One reason for having that added to /boot/grub/grub.conf is an fsck may first check "/" and then reboot before checking other file systems leaving them out and because partitions should not be checked when mounted so that leaves out "/" unless you reboot. Afterwards remove it from /boot/grub/grub.conf as servers don't need reboots often but you sure don't want to wait each reboot for a forced fsck to complete.



Your root file system is located at /dev/VolGroup00/LogVol00 and /tmp appears to be a file in the non-standard "/filesystems" directory. Unless loop-mounted file systems don't get mounted automagically there's nothing wrong with that. It's just not as optimal a configuration as it could be but that's your choice.

1 members found this post helpful.

Thanks, I've set the run level to 3.

I can try forcing fsck on boot, but I'm trying to minimise the risk of the server not coming back up as I have to ask a remote engineer to force the boot each time otherwise. So I'm just trying a couple of other things first, before I try that.

I actually tried yesterday to unmount some partitions to run fsck on them. I did /tmp and it was fine. Then I tried doing the same on /dev/shm and I got

Tried googling it but can't work out what this means. Do you know?

Also, if I understand correctly, I can do this on all partitions apart from / which can only be done in maintenance mode when booting or by forcing fsck to run automatically when booting. Is that right?

Also, how easy is it for me to add a seperate partition for /var now after install? Currently / is taking up most of the space. Would you recommend shrinking down / and then creating /var partition? According to the Plesk website /var is where all the diskspace gets used.

tmpfs is a virtual file system in RAM, not one you should fsck like your physical or LVM partitions.


Yes, any partition that hasn't got files open (everything is a "file" in Linux) you can't close.


I don't know what your definition of "easy" is. You'll have to shrink the existing LV, add a new one, create a file system, mount it somewhere, copy /var contents over, then mount it as (or over) /var when all files in /var are closed. That's why I suggested previously that if you thought your layout could be improved you could consider starting from scratch again...

1 members found this post helpful.

OK thanks for the advice.