[LU-9416] lmv_hsm_ct_register() should add kkuc before sending registration RPCs

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.11.0, Lustre 2.10.2
• Labels: hsm
• Severity: 3
• Assignee: Henri Doreau (Inactive)
• Reporter: John Hammond

Comments

[Gerrit Updater]

John L. Hammond ([e-mail]) merged in patch https://review.whamcloud.com/29687/
Subject: LU-9416 hsm: add kkuc before sending registration RPCs
Project: fs/lustre-release
Branch: b2_10
Current Patch Set:
Commit: 46aee8eba9000e458d231c975716feb275b8eb92

[Gerrit Updater]

Minh Diep ([e-mail]) uploaded a new patch: https://review.whamcloud.com/29687
Subject: LU-9416 hsm: add kkuc before sending registration RPCs
Project: fs/lustre-release
Branch: b2_10
Current Patch Set: 1
Commit: 45b51a053c014a11b49768757ab6d08103f8547e

[Peter Jones]

Landed for 2.11

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/28751/
Subject: LU-9416 hsm: add kkuc before sending registration RPCs
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: bdb04079570dd2de3b9310221e0206d64efb9ecc

[John Hammond]

Correct.

[Bruno Faccini (Inactive)]

So this should be kind of a follow-on to my previous patch for LU-9038, that fixed the crashes that could occur before and during the same race but introduced this retry mechanism from CDT, right...

[John Hammond]

> Could this be to avoid that some KUC msgs/requests from Server to Client could be lost, or to cause other inconvenient, during some racy cases ??

Correct. I have seen in sanity-hsm runs where the resistration completed and the CDTs sent HSM actions just before the kkuc registration happens. In this case the client drops the actions because there are no CT pipes in the KKUC list. Then the CDT must wait and resent the actions causing a 200+ second delay.

[Bruno Faccini (Inactive)]

John,
Can you at least comment/detail here about the main reason/need for this change ??!!
Could this be to avoid that some KUC msgs/requests from Server to Client could be lost, or to cause other inconvenient, during some racy cases ??... Or simply to do local operations first and thus prevent from doing useless remote/MDTs registrations in case of failure ?

[Gerrit Updater]

Henri Doreau ([e-mail]) uploaded a new patch: https://review.whamcloud.com/28751
Subject: LU-9416 hsm: add kkuc before sending registration RPCs
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: f76ade97e83d909f0247c93ae4e560ff2b442b0b