[LU-3004] MGS llog choke on to much parameter data

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.4.0
• Type: Bug
• Priority: Major
• Reporter: James A Simmons
• Assignee: Zhenyu Xu
• Resolution: Duplicate

Environment

Lustre 2.3.62 servers and clients. The system has 4 OSS with 28 real disk but I used lvm to create 56 OST per OSS thus slicing the disk into 8 partitions to use.

• Severity: 3
• Epic: mount

Description

In attempt to create a large stripe system with lvm I ran into a issue that prevented the MGT from mounting. The system has 4 OSS with 28 real disk but I used lvm to create 56 OST per OSS thus slicing the disk into 8 partitions to use. When I went to mount the file system the MGT refused to mount and gave the following error.

LustreError: 19657:0:(mgs_llog.c:1540:add_param()) params are too long:
failover.node=10.37.248.63@o2ib1 failover.node=10.37.248.64@o2ib1
failover.node=10.37.248.65@o2ib1 failover.node=10.37.248.62@o2ib1
failover.node=10.37.248.63@o2ib1 failover.node=10.37.248.64@o2ib1
failover.node=10.37.248.65@o2ib1 failover.node=10.37.248.62@o2ib1
failover.node=10.37.248.63@o2ib1 failover.node=10.37.248.64@o2ib1
failover.node=10.37.248.65@o2ib1 failover.node=10.37.248.62@o2ib1
failover.node=10.37.248.63@o2ib1 failover.node=10.37.248.64@o2ib1
failover.node=10.37.248.65@o2ib1 failover.node=10.37.248.62@o2ib1
failover.node=10.37.248.63@o2ib1 failover.node=10.37.248.64@o2ib1
failover.node=10.37.248.65@o2ib1 failover.node=10.37.248.62@o2ib1
failover.node=10.37.248.63@o2ib1 failover.node=10.37.248.64@o2ib1
failover.node=10.37.248.65@o2ib1 failover.node=10.37.248.62@o2ib1
failover.node=10.37.248.63@o2ib1 failover.node=10.37.248.64@o2ib1
failover.node=10.37.248.65@o2ib1 failover.node=10.37.248.62@o2ib1
failover.node=10.37.248
[ 7939.272546] LustreError: 19657:0:(mgs_llog.c:3557:mgs_write_log_target())
Can't write logs for lustre-MDT0000 (-22)
[ 7939.283312] LustreError: 19657:0:(mgs_handler.c:408:mgs_handle_target_reg())
Failed to write lustre-MDT0000 log (-22)
[ 8111.767428] Lustre: 13267:0:(client.c:1868:ptlrpc_expire_one_request()) @@@

Issue Links

• duplicates: LU-3038 OST nid is being messed up in the con...

Comments

[Zhenyu Xu]

patch tracking at http://review.whamcloud.com/5799

commit message

    LU-3004 mgs: don't add duplicated param
    
    mgs target info param does not need contain duplicated key val pair.
    It is the same case for lustre disk data params member.

[Zhenyu Xu]

dup of LU-3038