centos - UUID's, NFS, SunGridEngine =&...
A partition and disk imaging/cloning program
Brought to you by:
steven_shiau
Menu
▾
▴
Ok so kudos to clonezilla(CZ) devs. Ive used CZ live for some time and am now hoping to use it a bit differently. we have a SunGridEngine running on Centos 6.2 with 8 exec nodes. And id like to bump up all of our nodes to the latest version (6.2). However after cloneing a "gold image" from one of the boxes and restoring the 'gold' image on a different machine in the same cluster im having a few issues that i cant put my finger on.
summary:
(one 'gold' image -> several unique machines with identical hardware as the original machine used to build 'gold' image)
Ive :
1) changed the hostname
2) edit /etc/sysconfig/network
3) edit /etc/sysconfig/network-scripts/ifcfg* (etho, eth1)
4) checked /etc/fstab , /etc/httpd.conf.d, nuked the /tmp directory (removing any old stale nfs locks)
5) nuke and regenerate /etc/ssh keys
I am having issues mounting from my nfs server. the cloned machines are all clients, and the mounts do work, although sometimes I get a "mount.nfs: /shared is busy or already mounted" message. My SunGridEngine (SGE) binaries are mounted on an nfs share. Yet although my SGE, does recognize that all the nodes in the cluster are "there" - i cant launch any jobs on the boxes.. I was thinking that since clonezilla copied the UUID of the disks across my 8 nodes , perhaps my nfs server is somehow seeing that there is only one UUID and gets confused(???) - at least thats my hypothesis. I cant really explain it - perhaps others have some thoughts… but a cloned system only seems to work if i poweroff all the other nodes - making me think that something was left behind. I know windows has a 'sysprep' utility , and linux has a "sys-unconfig" - which i tried thinking it may fix some issues but no avail. I also tried changing the disk UUID on one of the boxes - in that senario was able to make the uuid different from the other cloned boxes but after the change it the box wouldnt boot…. so just restored the gold image again.
I was originally using CZ live (0922 natty) distro and decided to test the latest and see that there is a new "1-2-mdisks" and am wondering if i should have used this to restore the gold image to the other machines (intead of doing a plain restoredisk under natty)…. can anyone comment on this or has anyone encountered any issues. I know there is 'kickstarter' but would really like to see if i can get this cloning thing working… as the cloned machine is already tested , and configured -and with kickstart id have to select all the install packages and still retest to make sure all of our development programs still work.
if anyone can offer any ideas that would be terrific
(sorry for the long post)
"1-2-mdisks" is the same thing. It just restores one image to multiple disks at the same time. The mechanism is the same as you restore one by one.
And, yes, I think you are right. You should change the partition UUID.
Steven.