I've been trying to implement the quota system for a Linux cluster. I'm trying it out on a mini-cluster of 4 CentOS 6.3 VMs. I've named them:
lion-login lion-sn1 lion-cn01 lion-cn02
All nodes are on the virtual subnet 192.168.56.0/24. lion-sn1 (storage node 1) is the NFS server and the other three nodes are the NFS clients. /etc/exports on lion-sn1:
/home 192.168.56.0/24(rw)
/etc/fstab on the other nodes:
lion-sn1:/home /home nfs defaults,usrquota 0 2
User quotas are enabled on /home. rpc.rquotad is running on all 4 nodes.
I'm having two problems:
1. I've confirmed that /etc/init.d/halt is running 'quotaoff -aug' before the file systems are unmounted but the shutdown of lion-sn1 is usually delayed for half a minute because a process is supposedly using /home. Any suggestions on how I could troubleshoot this?
2. Quotas work fine on lion-sn1 (the NFS server). Here's what happens when I copy several large ISOs into a user's home directory:
2. Quotas work fine on lion-sn1 (the NFS server). Here's what happens when I copy several large ISOs into a user's home directory who's logged onto that server:
$ quota Disk quotas for user ulrick (uid 1000): Filesystem blocks quota limit grace files quota limit grace /dev/sdc1 68896 921600 1048576 3147 0 0 $ cp -v /mnt/home/images/CentOS-6.* . `/mnt/home/images/CentOS-6.2-i386-netinstall.iso' -> `./CentOS-6.2-i386-netinstall.iso' `/mnt/home/images/CentOS-6.3-i386-bin-DVD1.iso' -> `./CentOS-6.3-i386-bin-DVD1.iso' sdc1: warning, user block quota exceeded. sdc1: write failed, user block limit reached. cp: writing `./CentOS-6.3-i386-bin-DVD1.iso': Disk quota exceeded `/mnt/home/images/CentOS-6.3-x86_64-bin-DVD1.iso' -> `./CentOS-6.3-x86_64-bin-DVD1.iso' cp: writing `./CentOS-6.3-x86_64-bin-DVD1.iso': Disk quota exceeded $ quota Disk quotas for user ulrick (uid 1000): Filesystem blocks quota limit grace files quota limit grace /dev/sdc1 1048576* 921600 1048576 7days 3150 0 0
On the NFS client nodes, the attempt to use space beyond the hard quota limit is prevented but no warning is given for exceeding the soft limit, and the message issued for the hard limit excession is 'Input/output error' rather than 'Disk quota exceeded'. Note that the 'quota' command seems to be working correctly:
$ quota Disk quotas for user ulrick (uid 1000): Filesystem blocks quota limit grace files quota limit grace lion-sn1:/home/ 68896 921600 1048576 3147 0 0 $ cp -v /mnt/home/images/CentOS-6.* . `/mnt/home/images/CentOS-6.2-i386-netinstall.iso' -> `./CentOS-6.2-i386-netinstall.iso' `/mnt/home/images/CentOS-6.3-i386-bin-DVD1.iso' -> `./CentOS-6.3-i386-bin-DVD1.iso' cp: writing `./CentOS-6.3-i386-bin-DVD1.iso': Input/output error `/mnt/home/images/CentOS-6.3-x86_64-bin-DVD1.iso' -> `./CentOS-6.3-x86_64-bin-DVD1.iso' cp: writing `./CentOS-6.3-x86_64-bin-DVD1.iso': Input/output error $ quota Disk quotas for user ulrick (uid 1000): Filesystem blocks quota limit grace files quota limit grace lion-sn1:/home/ 1048576* 921600 1048576 6days 3150 0 0
Instead of the 'Input/output error', I'd like the user on the NFS client systems to see the 'Disk quota exceeded' messages they would get if they were logged on to the NFS server. Any suggestions on how I might resolve this issue?
Thanks, Dave