[LU-4167] Interop 2.4.1<->2.5 failure on test suite conf-sanity test_32d: unknown param max_dirty_mb

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.5.0, Lustre 2.5.1
• Fix Version/s: Lustre 2.6.0, Lustre 2.7.0, Lustre 2.5.3
• Type: Bug
• Priority: Minor
• Reporter: Maloo
• Assignee: Emoly Liu
• Resolution: Fixed

Environment

server: 2.4.1 RHEL6 ldiskfs
client: lustre-b2_5 build #2 RHEL6 ldiskfs

• Severity: 3

Description

This issue was created by maloo for sarah <[e-mail]>

This issue relates to the following test suite run: http://maloo.whamcloud.com/test_sets/75f79ad4-3eb9-11e3-a21b-52540035b04c.

The sub-test test_32d failed with the following error:

test_32d failed with 1

MDS console:

23:22:48:Lustre: 5409:0:(obd_mount.c:837:lustre_check_exclusion()) Excluding t32fs-OST0000-osc (on exclusion list)
23:22:48:LustreError: 5409:0:(obd_config.c:1303:class_process_proc_param()) t32fs-OST0000-osc: unknown param max_dirty_mb=15
23:22:48:Lustre: DEBUG MARKER: /usr/sbin/lctl conf_param t32fs-MDT0000.lov.stripesize=4M

Comments

[Andreas Dilger]

It isn't possible to run LFSCK on an OST in Lustre 2.4. This test needs to be changed to skip running LFSCK for versions older than 2.4.93 or so.

[Emoly Liu]

http://review.whamcloud.com/8132

[Jian Yu]

Lustre client build: http://build.whamcloud.com/job/lustre-b2_5/5/
Lustre server build: http://build.whamcloud.com/job/lustre-b2_4/70/ (2.4.2)

The same failure occurred:
https://maloo.whamcloud.com/test_sets/73e57c7e-74f7-11e3-95ae-52540035b04c

[Bob Glossman (Inactive)]

seen again without interop:
https://maloo.whamcloud.com/test_sets/e694ab82-9545-11e3-80d2-52540035b04c

[Jian Yu]

Lustre client build: http://build.whamcloud.com/job/lustre-b2_5/39/ (2.5.1 RC1)
Lustre server build: http://build.whamcloud.com/job/lustre-b2_4/70/ (2.4.2)

The same failure occurred:
https://maloo.whamcloud.com/test_sets/416da544-a54b-11e3-9fee-52540035b04c

[Jodi Levi (Inactive)]

Patch landed to Master.

[Jian Yu]

Lustre client build: http://build.whamcloud.com/job/lustre-b2_5/61/
Lustre server build: http://build.whamcloud.com/job/lustre-b2_4/73/ (2.4.3)

The same failure occurred: https://maloo.whamcloud.com/test_sets/7e385b20-ead5-11e3-966a-52540035b04c

[Jian Yu]

Lustre client build: https://build.hpdd.intel.com/job/lustre-b2_5/80/
Lustre server build: http://build.whamcloud.com/job/lustre-b2_4/73/ (2.4.3)

The same failure still occurred:
https://testing.hpdd.intel.com/test_sets/d0ffc616-266e-11e4-8ee8-5254006e85c2

Lustre b2_5 build #80 contains patch http://review.whamcloud.com/8132.

Hi Emoly, could you please take a look at this issue? Thanks!

[Emoly Liu]

Hi Fanyong, could you please help this one? Thanks
I find this error happened even during patch set 8 maloo test in http://review.whamcloud.com/8132.

"Patch Set 8:
It is very strange failure:
00:08:45:Lustre: DEBUG MARKER: /usr/sbin/lctl lfsck_start -M t32fs-OST0000 00:08:45:LustreError: 2706:0:(ofd_obd.c:1568:ofd_iocontrol()) t32fs-OST0000: not supported cmd = -1073191194 00:09:26:Lustre: DEBUG MARKER: /usr/sbin/lctl mark conf-sanity test_32d: @@@@@@ FAIL: Start OI scrub on OST0
Means ofd_iocontrol() cannot recognise the @cmd OBD_IOC_START_LFSCK (-1073191194).
So unless the test source code is very old (such as 2.4), otherwise, the 2.6 candidate should has processed it properly. I want to check the built source code (#20521), but I cannot find it..."

[nasf (Inactive)]

You can directly try LFSCK on the specified server build to check whether such version support OI scrub on OST or not.

lctl lfsck_start -M ${fsname}-OST0000

[Emoly Liu]

Thanks, Fanyong. I will try it.

[Emoly Liu]

I ran the command "lfsck_start" on 2.4.3 server like fanyong suggested. It reported the same error

[root@onyx-25 ~]# cat /proc/fs/lustre/version
lustre: 2.4.3
kernel: patchless_client
build:  2.4.3-RC1--PRISTINE-2.6.32-358.23.2.el6_lustre.x86_64
[root@onyx-25 ~]# lctl lfsck_start -M lustre-OST0000
Fail to start LFSCK: Inappropriate ioctl for device
[root@onyx-25 ~]# tail /var/log/messages
...
Aug 21 00:30:21 onyx-25 kernel: LustreError: 10157:0:(ofd_obd.c:1568:ofd_iocontrol()) lustre-OST0000: not supported cmd = -1073191194

So, let me add OST version check into that script to fix the problem.

[Andreas Dilger]

It would be great to also fix the ofd_iocontrol()message to print the ioctl() CMD argument as hex instead of a signed integer.

[Emoly Liu]

The patch for master is at http://review.whamcloud.com/11556
The patch for b2_5 is at http://review.whamcloud.com/11574