In addition to the simple fix (s/ENOENT/-ENOENT/) mentioned after my
previous post, I've also discovered and fixed a nasty semaphore deadlock
condition.
I've run some mild tests and found that I can successfully:
- Create/mount ext3 file system on node-A
- Create cluster snapshot
- Mount snapshot on node-B for reading, while node-A is untarring linux
kernels
I'm including a script, 'clusterize_snapshot.sh', which converts
snapshots to cluster-aware snapshots. The steps I perform are as
follows:
1) on one node, create a snapshot ('lvcreate -s ...')
2) 'vgchange -ay' on all nodes (you must be using shared storage)
3) run 'clusterize_snapshot.sh' on all nodes
brassow
Attachment:
clusterize_snapshot.sh
Description: application/shellscript