[LU-6086] verify MDTs are running the same Lustre version

• Type: Improvement
• Resolution: Fixed
• Priority: Major
• Fix Version/s: Lustre 2.7.0, Lustre 2.5.5
• Affects Version/s: Lustre 2.7.0, Lustre 2.5.5
• Assignee: nasf (Inactive)
• Reporter: Andreas Dilger

Description

We require that MDTs and OSTs be running the same version of Lustre (i.e. no interoperability between MDTs) for DNE and LFSCK. While we always test this way and know this requirement ourselves, I know that there will be users who don't know this in the future, and I don't think this is actually checked or enforced anywhere.

The MDTs and OSTs should check at connect time via ocd_version that the connecting server is running the same major version of Lustre (e.g. 2.5.x or 2.7.x). Otherwise we may run into interoperability issues if someone is trying to run different versions of Lustre on the MDTs (e.g. DNE1 2.5 - DNE2 2.6 , or DNE2 2.6 - DNE2 2.8 async_commit, or LFSCK changes like http://review.whamcloud.com/13172 or similar).

Issue Links

• is related to: LU-6307 Interop 2.6.0<->2.7 recovery-small test_105: MGS refused the connection from different version MDT

Comments

[Joseph Gmitter (Inactive)]

Reopening/resolving to be able to adjust the Fix Version. This was entirely landed for 2.7. There is nothing landed here that was addressed in 2.8.

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/13285/
Subject: LU-6086 obdclass: check peer's version for MDT-MDT connection
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 0030ed640284e4ed5afa9187c5a907141132b083

[nasf (Inactive)]

I would say that we will only support the rolling upgrade between small versions, such as the patch version difference is within 3. But if rolling upgrade happens between major versions, such as from Lustre-2.6 to Lustre-2.7, it will get failure.

Andreas, what's your thought for that?

[Oleg Drokin]

I worry about rolling upgrades that could be one mds or one ost at a time

[Gerrit Updater]

Fan Yong ([e-mail]) uploaded a new patch: http://review.whamcloud.com/13287
Subject: LU-6086 obdclass: check peer's version for server connection
Project: fs/lustre-release
Branch: b2_5
Current Patch Set: 1
Commit: 9621c9fc081bf5a3866c188c3a4bd4de51a6e930

[Gerrit Updater]

Fan Yong ([e-mail]) uploaded a new patch: http://review.whamcloud.com/13285
Subject: LU-6086 obdclass: check peer's version for server connection
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 3ccbc301d2f987c6befed61459d6bcef9fc2b53a

[nasf (Inactive)]

If the OST found that the MDT is running different version, should we forbid the connection? or forbid some LFSCK operations? Or just print some warning message on console?

On the other hand, I am not sure whether it is part of the roll upgrade/downgrade among servers.