[LU-8404] When service node nid is incorrect, MDT log message missing bad nid

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.7.0
• Type: Bug
• Priority: Minor
• Reporter: Gary Hagensen (Inactive)
• Assignee: Hongchao Zhang
• Resolution: Duplicate
• Severity: 3

Description

When entering a bad nid address in the --servicenode argument. This caused the MDT to not be able to establish a connection and reports in the log as below.

Note the 0@<0:0> at the end that, I am guessing, should be the servicenode failover args and is reporting "0:0" instead. If it had reported the correct args, this would have been a much easier problem to figure out.

Also in the mds log:

class_config_llog_handler()) MGC192.168.20.121@o2ib: cfg command failed: rc = -2
cmd=cf003 0:lustrefs-OST0010-osc-MDT0000  1:lustrefs-OST0010_UUID  2:0@<0:0>

Comments

[Andreas Dilger]

Gary, can you please provide more details about how the NID was incorrect, so that this can be debugged/fixed properly and there can be a test case added for this. As it is, I have no idea how to reproduce this problem.

[Gary Hagensen (Inactive)]

The bad servicenode argument I used was the node's management interface @ tcp0 . So it was a valid IP, but not a valid NID as it is not on an LNET. The LNET interface was the IB interface. So I assume to reproduce all you would need is a system (or VMs) with 2 interfaces on the HA pair, one for management, one for LNET and define the service nodes to be on the non-lnet interface.

[Hongchao Zhang]

the patch http://review.whamcloud.com/19933/ from LU-8091 can fix the problem.

[Peter Jones]

So this is confirmed as a duplicate of LU-8091