Details
-
Bug
-
Resolution: Fixed
-
Minor
-
Lustre 2.1.0, Lustre 2.2.0, Lustre 2.3.0, Lustre 2.1.2
-
Scientific Linux 5.5 (servers), Scientific Linux 5.5 & 6.2 (clients)
-
3
-
4459
Description
When upgrading from Lustre 1.8 to 2.1, if the parameter rootsquash is set, Lustre throws an error and denies mounting the filesystem. This is the error from the MDS logs:
Mar 7 20:03:35 mds01 kernel: LustreError: 6610:0:(obd_config.c:1140:class_process_proc_param()) home-MDT0000: unknown param rootsquash=99:99
Mar 7 20:03:35 mds01 kernel: LustreError: 6610:0:(obd_config.c:1361:class_config_llog_handler()) Err -38 on cfg command:
Mar 7 20:03:35 mds01 kernel: Lustre: cmd=cf00f 0:home-MDT0000 1:mdt.rootsquash=99:99
Mar 7 20:03:35 mds01 kernel: LustreError: 15c-8: MGC172.16.4.3 at tcp: The configuration from log 'home-MDT0000' failed (-38). This may be the result of communication errors between this node and the MGS, a bad configuration, or other errors. See the syslog for more information.
Mar 7 20:03:35 mds01 kernel: LustreError: 6124:0:(obd_mount.c:1192:server_start_targets()) failed to start server home-MDT0000: -38
Mar 7 20:03:35 mds01 kernel: LustreError: 6124:0:(obd_mount.c:1723:server_fill_super()) Unable to start targets: -38
Mar 7 20:03:35 mds01 kernel: Lustre: Failing over home-MDT0000
We were able to overcome this using tunefs.lustre with the option --writeconf added. This should be managed automatically (ideally, migrating the configuration) or at least, documented.
Attachments
Issue Links
- is related to
-
-
- Resolved
-
- Trackbacks
-
![[Confluence: Whamcloud Community Space]](/images/icons/generic_link_16.png "[Confluence: Whamcloud Community Space]")
Changelog 2.1
Changes from version 2.1.2 to version 2.1.3 Server support for kernels: 2.6.18308.13.1.el5 (RHEL5) 2.6.32279.2.1.el6 (RHEL6) Client support for unpatched kernels: 2.6.18308.13.1.el5 (RHEL5) 2.6.32279.2.1....
Activity
Patch for b2_1 branch: http://review.whamcloud.com/3900
The one for master branch has been landed and also cherry-picked on b2_3 branch.
The remaining work on this ticket is to improve the patch on master branch to handle more parameters as per the suggestion from Andreas.
Patch for master branch: http://review.whamcloud.com/3836
The patches have been landed on master and b2_1 branches.
The remaining work on this ticket is to improve the patch on master branch to handle more parameters as per the suggestion from Andreas on http://review.whamcloud.com/3237 :
File lustre/mdt/mdt_handler.c
Line 4779: strlen(CFG_NOSQUASH_NID_OLD)) == 0 &&
If this list gets any longer (which I suspect it will, due to Di's patch http://review.whamcloud.com/1418 to split the MDT thread handling) then this should be modified to use an array of { "old", "new" } parameters, and if "new" is NULL then the parameter is obsolete and discarded. Then, the above code will just be a loop comparing the parameter against each "old" entry, and ending the loop if it matched, or "old" is NULL.I don't insist on this today, but since the code is already being changed, and will need to change again in the near future, it would be better to just get it right for the long term.
Patch for master branch: http://review.whamcloud.com/2680
The above patch was landed on master branch.
Here is the patch for b2_1 branch: http://review.whamcloud.com/3237.
The old "mdt.nosquash_nid" parameter also needs to be renamed to "mdt.nosquash_nids".
I agree that this should be handled internally to the MDT code in mdt_process_config() by renaming the parameter. The new parameter is called "mdt.root_squash" (note underscore).
This was done similarly for the quota configuration in LU-110 (http://review.whamcloud.com/354).
Please reopen ticket if there is additional work needed.