[LU-9943] LU-7124 caused a connection problems under load.

• Type: Bug
• Resolution: Fixed
• Priority: Critical
• Fix Version/s: Lustre 2.11.0
• Affects Version/s: Lustre 2.8.0, Lustre 2.9.0, Lustre 2.10.0
• Labels: lnet
• Severity: 3
• Assignee: Amir Shehata (Inactive)
• Reporter: Alexey Lyashkov

Description

LU-7124 is completely incorrect patch.
It decrease a WR array into kernel, but o2ib lnd stay assume about own queue depth.
It caused situation

Aug  8 21:53:00 lstr1n07 kernel: mlx5_warn:mlx5_0:mlx5_ib_post_send:4184:(pid 26272): 
Failed to prepare WQE
Aug  8 21:53:00 lstr1n07 kernel: mlx5_warn:mlx5_0:begin_wqe:4085:(pid 9590): work queue overflow

after several ENOMEM hits.