[LU-19746] Phantom connections cause replace_nids to fail

• Type: Improvement
• Resolution: Unresolved
• Priority: Major
• Fix Version/s: Lustre 2.18.0
• Affects Version/s: Lustre 2.18.0
• Labels: patch
• Severity: 3

Epic

server

• Assignee: Alexander Boyko
• Reporter: Alexander Boyko

Description

There's an issue during NID replacement at the customer site. With thousands of clients, unmounting all connections is difficult. When the MGS starts with the nosvc option, some clients reconnect, preventing the replace_nid operation from completing. Let's introduce a new mount option "noclient" to block all non-local connections for MGS, allowing replace_nid to succeed.

.....mgs_replace_nids()) Only MGS is allowed to be started
.....mgs_iocontrol()) MGS: error replacing nids: rc = -115

Issue Links

• is related to: LU-19834 conf-sanity test_162: @@@@@@ FAIL: replace nids failed