-
Task
-
Resolution: Unresolved
-
Major
-
None
-
None
-
None
-
2
-
rhel-ha
-
HA-PCMK Sprint #3: 2025-10-13, HA-PCMK Sprint #5
-
2
-
False
-
-
None
What were you trying to do that didn't work?
Adding/removing devices to fence_scsi & fence_mpath restart other cluster resources, which shouldn't happen - see https://bugzilla.redhat.com/show_bug.cgi?id=1872376 & https://bugzilla.redhat.com/show_bug.cgi?id=2177996.
Please provide the package NVR for which the bug is seen:
found in pacemaker-2.1.9-1.el9 (RHEL9.6), but goes back to pacemaker-2.1.7-1.el9.x86_64 (RHEL9.4). The issue is not present in pacemaker-2.1.6-10.1.el9_3.x86_64 (RHEL9.3).
How reproducible is this bug?:
always
Steps to reproduce
Have a cluster with shared devices, fence_scsi a some other resources, check time of start operation for the resources, then update a device for fence_scsi and check the start operation again.
[root@virt-535 ~]# ls -lr /dev/disk/by-id/ | grep -m 3 "sda\|sdb\|sdc"
lrwxrwxrwx. 1 root root 9 Dec 4 10:45 wwn-0x600140583126160982e4b92b7a8035fd -> ../../sdb
lrwxrwxrwx. 1 root root 9 Dec 4 10:45 wwn-0x6001405388343a967d54aada774c632f -> ../../sda
lrwxrwxrwx. 1 root root 9 Dec 4 10:45 wwn-0x6001405027be43412bf4c2989c61174e -> ../../sdc
[root@virt-535 ~]# pcs stonith create scsi-fencing fence_scsi devices="/dev/disk/by-id/wwn-0x600140583126160982e4b92b7a8035fd" pcmk_host_check="static-list" pcmk_host_list="virt-535 virt-536" pcmk_reboot_action="off" meta provides="unfencing"
[root@virt-535 ~]# pcs resource create r1 ocf:heartbeat:Dummy
[root@virt-535 ~]# crm_resource --list-all-operations --resource r1 | grep start
r1 (ocf:heartbeat:Dummy): Started: r1_start_0 (node=virt-536, call=24, rc=0, last-rc-change='Wed Dec 4 15:14:12 2024', exec=19ms): complete
> start time is at 15:14:12.
[root@virt-535 ~]# pcs stonith update-scsi-devices scsi-fencing add /dev/disk/by-id/wwn-0x6001405388343a967d54aada774c632f
[root@virt-535 ~]# crm_resource --list-all-operations --resource r1 | grep start
r1 (ocf:heartbeat:Dummy): Started: r1_start_0 (node=virt-536, call=28, rc=0, last-rc-change='Wed Dec 4 15:15:22 2024', exec=17ms): complete
> new start time is 15:15:22.
Expected results
Other cluster resources are not restarted when editing scsi devices for fence_scsi & fence_mpath.
Actual results
The resources are restarted after editing scsi devices.
Attachment
4 CIB files - two from RHEL9.4, before updating the devices for fence_scsi and after. Another two for comparison from version RHEL9.3 (also before and after the disks update), where the problem wasn't present.
After investigation done by mlisik@redhat.com, it seems that the problem might point to change of value of
<nvpair id="status-1-.node-unfenced" name="#node-unfenced" value="1733331940"/>
This can be seen in the diff between RHEL9.4-before.xml and RHEL9.4-after.xml.
- is blocked by
-
RHEL-70283 Adding scsi device to a cluster results in a restart of cluster resources
-
- In Progress
-
