Bug 175158
| Summary: | scsi devices busy in cluster | ||
|---|---|---|---|
| Product: | [Retired] Red Hat Cluster Suite | Reporter: | Nate Straz <nstraz> |
| Component: | lvm2-cluster | Assignee: | Christine Caulfield <ccaulfie> |
| Status: | CLOSED NOTABUG | QA Contact: | Cluster QE <mspqa-list> |
| Severity: | medium | Docs Contact: | |
| Priority: | medium | ||
| Version: | 4 | CC: | agk |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | i386 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2005-12-07 21:42:30 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
|
Description
Nate Straz
2005-12-06 23:52:02 UTC
Tried it with DLM on the same cluster and ran into the same problem. I have another cluster running the same tests without problems. Same storage hardware and HBAs too. I did some digging with systemtap and found why EBUSY is being returned.
In rescan_partitions(), bdev->bd_part_count == 1 on the failing nodes.
Here is the systemtap script I used:
/* rescan.stp
This is a simple script to figure out what is happening with block device
rescans.
*/
global count_rescan
probe kernel.function("blkdev_ioctl")
{
log("blkdev_ioctl cmd " . string($cmd));
}
probe kernel.function("blkdev_reread_part")
{
log("entered blkdev_reread_part");
}
probe kernel.function("rescan_partitions")
{
log("entered rescan_partitions");
log("part_count = " . string($bdev->bd_part_count));
}
probe kernel.function("invalidate_partition")
{
log("entered invalidate_partition");
}
probe begin { log("starting probe") }
probe end
{
log("ending probe");
}
Passing output:
[root@morph-01 ~]# stap /tmp/rescan.stp
starting probe
blkdev_ioctl cmd 21286
blkdev_ioctl cmd 21286
blkdev_ioctl cmd 4703
entered blkdev_reread_part
entered rescan_partitions
part_count = 0
entered invalidate_partition
blkdev_ioctl cmd 4712
blkdev_ioctl cmd -2147216782
blkdev_ioctl cmd 21286
ending probe
Failing output:
[root@morph-02 ~]# stap /tmp/rescan.stp
starting probe
blkdev_ioctl cmd 21286
blkdev_ioctl cmd 21286
blkdev_ioctl cmd 4703
entered blkdev_reread_part
entered rescan_partitions
part_count = 1
blkdev_ioctl cmd 21286
ending probe
Ah ha! I found the holder! On the nodes that weren't working, there was still an entry in the DM table. After using dmsetup to remove it, I was back in business. ARG! Found that lvm.conf was wrong on one node. It didn't have cluster locking enabled. |