[LU-13028] LNet Discovery: toggling discovery on/off is not handled properly

• Status: Resolved
• Project: Lustre
• Fix Version/s: Lustre 2.14.0
• Type: Bug
• Priority: Minor
• Reporter: Amir Shehata (Inactive)
• Assignee: Amir Shehata (Inactive)
• Resolution: Fixed
• Severity: 3

Description

• When a peer goes from discovery on to discovery off, it's possible that its NIs change, such that the list of NIs local to the node are no longer valid.
  • Before turning off discovery send a broadcast to all peers with the new status
  • When a node receives a peer's state with discovery off. It should just delete the existing peer
    • If the peer is being used for as a gateway, then allocate a new peer for the route
  • When sending to a peer fails, it should be flagged for rediscovery

Issue Links

• is related to: LU-12312 sanity-sec: test_31: 'network' mount ...

Comments

[Gerrit Updater]

Amir Shehata ([e-mail]) uploaded a new patch: https://review.whamcloud.com/36919
Subject: LU-13028 lnet: advertise discovery off
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 51da7585f1fb1b2c74f93300c23d8d4e36b28093

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/36919/
Subject: LU-13028 lnet: advertise discovery when toggled
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 4577410165641e3756406aca7f9a21c73d1fd630

[Peter Jones]

Landed for 2.14

[Andreas Dilger]

The patch https://review.whamcloud.com/36919 "LU-13028 lnet: advertise discovery when toggled" changed lnet, but was submitted with "trivial" and looks like it is causing sanity-sec test_31 to fail 100% of the time with:

LNet Dynamic Peer Discovery is enabled on this node. 
'network' mount option cannot be taken into account.

I've pushed a revert patch to see if this change is the source of the problem.