[LU-3546] 2.4.0<->2.3 interop: sanity test 56w: MDS unknown opcode 61

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.4.1, Lustre 2.5.0
• Affects Version/s: Lustre 2.4.0
• Labels: mn1 yuc2

Environment

Lustre server build: https://build.whamcloud.com/job/lustre-chris/25/
Lustre client build: http://build.whamcloud.com/job/lustre-b2_4/13/

• Severity: 3
• Assignee: Jinshan Xiong (Inactive)
• Reporter: Jian Yu

Description

sanity test 56w hung as follows:

== sanity test 56w: check lfs_migrate -c stripe_count works == 00:42:51 (1372405371)
/usr/bin/lfs_migrate -y -c 6 /mnt/lustre/d0.sanity/d56w/file1

Console log on the MDS showed that:

00:39:22:Lustre: DEBUG MARKER: == sanity test 56w: check lfs_migrate -c stripe_count works == 00:42:51 (1372405371)
00:39:44:LustreError: 14887:0:(mdt_handler.c:3026:mdt_msg_check_version()) MDS unknown opcode 61
00:39:44:LustreError: 14887:0:(mdt_handler.c:3065:mdt_handle0()) mdt drops mal-formed request
00:40:25:Lustre: lustre-MDT0000: Client ca0ab3a3-a02f-6aeb-30e4-9241dc18b04d (at 10.10.4.121@tcp) reconnecting
00:40:25:LustreError: 22982:0:(mdt_handler.c:3026:mdt_msg_check_version()) MDS unknown opcode 61
00:40:25:LustreError: 22982:0:(mdt_handler.c:3065:mdt_handle0()) mdt drops mal-formed request
00:41:07:Lustre: lustre-MDT0000: Client ca0ab3a3-a02f-6aeb-30e4-9241dc18b04d (at 10.10.4.121@tcp) reconnecting
00:41:07:LustreError: 14887:0:(mdt_handler.c:3026:mdt_msg_check_version()) MDS unknown opcode 61
00:41:07:LustreError: 14887:0:(mdt_handler.c:3065:mdt_handle0()) mdt drops mal-formed request
00:41:49:Lustre: lustre-MDT0000: Client ca0ab3a3-a02f-6aeb-30e4-9241dc18b04d (at 10.10.4.121@tcp) reconnecting

Maloo report: https://maloo.whamcloud.com/test_sets/0219094e-e0f0-11e2-b3fd-52540035b04c

Comments

[Sarah Liu]

Also hit this problem when testing interop between 2.3.0 server and 2.5 client:

https://maloo.whamcloud.com/test_sets/4926deda-02b8-11e3-a4b4-52540035b04c

[Jodi Levi (Inactive)]

Can this be closed? Or is there more work to do?

[Oleg Drokin]

cherrypicked compile fix to b2.1

[Jodi Levi (Inactive)]

The fix for the build failure needs to be cherry picked to b2_1

[Jodi Levi (Inactive)]

Can this ticket be closed now? Or does the patch still need to be cherry picked?

[Oleg Drokin]

the patch was landed and cherry-picked to b2_1 and b2_4

[Andreas Dilger]

This patch is also needed for 2.1.7.

[Jinshan Xiong (Inactive)]

patch is at: http://review.whamcloud.com/6864, the patch is for master but we should cherry-pick it to b2_3 and b2_4.

[Emoly Liu]

The following opcodes are not supported by 2.3 MDS.

        MDS_HSM_STATE_GET       = 54,
        MDS_HSM_STATE_SET       = 55,
        MDS_HSM_ACTION          = 56,
        MDS_HSM_PROGRESS        = 57,
        MDS_HSM_REQUEST         = 58,
        MDS_HSM_CT_REGISTER     = 59,
        MDS_HSM_CT_UNREGISTER   = 60,
        MDS_SWAP_LAYOUTS        = 61,

[Andreas Dilger]

The "lfs migrate" RPC is not supported by the 2.3 MDS, but I guess it is a bug in the 2.x MDS code that it just drops this reply instead of returning an error to the caller? It should definitely return -EPROTO or -EOPNOTSUPP or similar.