Details
-
Bug
-
Resolution: Fixed
-
Minor
-
None
-
None
-
3
-
9223372036854775807
Description
The OBD may not setup successful while onu_upcall be invoked. So we have to check it in cl_ocd_update(). Since customer doing writeconf after this issue. So we have not possible to figure out why OBD was not setup. Below is original report from customer:
Hello ,
MDT (ldiskfs, LSI 9286 syncro)
OST (openzfs 0.6.4.1-1, JBOD)
We add some HDDs to JBOD enclosure.
1. Update /etc/zfs/vdev_id.conf
2. Create zpool
3. Enable compression
4. mkfs.lustre (ost-24,ost-25)
5. export zpool
6. Add zfs and lustre mount resource to pacemaker
All operates are successful
A few minutes later,
7. MDT node reboot, I don 't know why.
8. Login MDT passive node , pcs show MDT mount fail
9. All lustre clients kernel panic(oh ,shit)
10. e2fsck mdt
11. writeconf mdt and osts
12. mount mgt mdt and osts
14. lustre client mount failed
15. remount mdt and osts, mdt mount failed
16. writeconf
17. mount ost-0, lustre client mount successful
18. umount ost-0
19. mount ost-24, lustre client mount failed
20. mdt could not remount
21. writeconf
22. exclude ost-24 and ost-25 , mount mdt and the other osts
23. mount mdt and osts
24. lustre client mount successful