lvm commands hangs when one leg of a mirrored lvm is removed in stretch cluster
Issue
- Testing lvm mirror on a 2 node stretch cluster running RHEL 5.5. Simulating a data center disaster, removing one leg to see how it behaves. I was expecting the lv to convert linear but it didn't happen. All lvm commands or commands on the file system hang.
Below 2 tests are performed:
- Removing luns
echo 1 > /sys/block/$path/device/delete. After removing luns, any i/o operation hangs (lvm commands, "ls" on the filesystem,...) - Forcing paths to be down
echo offline > /sys/block/$path/device/state. lv converts to linear as expected.
- We have 1 mirror with LVM with one disk in a site and another disk in other site. With this mirror we have a File System and the problem is that when we have a problem in one disk, the File System have a problem and is impossible have access to it
- 2 cluster nodes are located in 2 different places with shared SAN storage attached, one SAN at each site. The LVM mirror configured such that, it have one leg at one site and other leg on other site.
- Now, if storage connection from one site goes down, the filesystem is not accessible on either of node?
Environment
- Red Hat Enterprise Linux (RHEL) 5
- Logical Volume Manager (LVM) 2
- Device Mapper (DM) multipath
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase of over 48,000 articles and solutions.
Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.
