[LU-4374] lctl set_param -d -P mdt.<FSname>-MDT0000.hsm_control, returns a false error when default already set

• Type: Bug
• Resolution: Fixed
• Priority: Major
• Fix Version/s: Lustre 2.6.0
• Affects Version/s: Lustre 2.5.0, Lustre 2.6.0
• Severity: 3
• Assignee: Bruno Faccini (Inactive)
• Reporter: Bruno Faccini (Inactive)

Description

On a MDS :

[root@mds ~]# dmesg -c > /dev/null
[root@mds ~]# lctl set_param -P mdt.lustre-MDT0000.hsm_control=enabled
[root@mds ~]# dmesg
Lustre: Setting parameter general.mdt.lustre-MDT0000.hsm_control in log params
Lustre: MGS: non-config logname received: params
Lustre: Skipped 13 previous similar messages
LustreError: 3043:0:(mdt_coordinator.c:950:mdt_hsm_cdt_start()) lustre-MDT0000: Coordinator already started
[root@mds ~]# lctl set_param -d -P mdt.lustre-MDT0000.hsm_control
[root@mds ~]# dmesg
Lustre: Setting parameter general.mdt.lustre-MDT0000.hsm_control in log params
Lustre: MGS: non-config logname received: params
Lustre: Skipped 13 previous similar messages
LustreError: 3043:0:(mdt_coordinator.c:950:mdt_hsm_cdt_start()) lustre-MDT0000: Coordinator already started
Lustre: Disabling parameter general.mdt.lustre-MDT0000.hsm_control in log params
[root@mds ~]# lctl set_param -d -P mdt.lustre-MDT0000.hsm_control
error: executing set_param: Success
[root@mds ~]# dmesg
Lustre: Setting parameter general.mdt.lustre-MDT0000.hsm_control in log params
Lustre: MGS: non-config logname received: params
Lustre: Skipped 13 previous similar messages
LustreError: 3043:0:(mdt_coordinator.c:950:mdt_hsm_cdt_start()) lustre-MDT0000: Coordinator already started
Lustre: Disabling parameter general.mdt.lustre-MDT0000.hsm_control in log params
Lustre: Disabling parameter general.mdt.lustre-MDT0000.hsm_control in log params
LustreError: 3051:0:(mgs_handler.c:744:mgs_iocontrol()) MGS: setparam err: rc = 1
[root@mds ~]# 

This seem to come from the wrong interpretation of rc=1 being returned from mgs_modify() when no config-llog record modification occured (!rc && !mml->mml_modified) in llog_process().

Will provide a patch soon.

Comments

[Jodi Levi (Inactive)]

Patch landed to Master.

[Bruno Faccini (Inactive)]

Patch re-based and latest comments addressed in patch-set 6 of http://review.whamcloud.com/8567.

[Bruno Faccini (Inactive)]

Reviewers ok, waiting for patch to land now.

[Bruno Faccini (Inactive)]

New patch-set #2 submitted to fix issue but also address comments from J-C.

[Bruno Faccini (Inactive)]

Patch local testing does not show any more false error/msg when no config-llog update occurs because default/value is already set.

[Bruno Faccini (Inactive)]

Issue has been discovered during LU-4065 analysis.
Patch is at http://review.whamcloud.com/8567.