AnsweredAssumed Answered

RedHat multipath.conf Question.

Question asked by Wanjong Kim on Aug 30, 2018

Hello.

I have trouble with 16Gb SFP on G800. 3 times a month.

SFP failure symptom is a little wired.

RedHat 7.2 detects path error. but multipath on RedHat do not shutdown path(Approx. 1 hour and 30 minutes)

=================================

Aug 13 04:58:57 wel kernel: qla2xxx [0000:05:00.1]-801c:2: Abort command issued nexus=2:0:69 -- 1 2002.

Aug 13 04:58:57 wel kernel: sd 2:0:0:69: [sdei] tag#4 FAILED Result: hostbyte=DID_ERROR driverbyte=DRIVER_OK

Aug 13 04:58:57 wel kernel: sd 2:0:0:69: [sdei] tag#4 CDB: Read(10) 28 00 08 1a bc 00 00 04 00 00

Aug 13 04:58:57 wel kernel: blk_update_request: I/O error, dev sdei, sector 135969792

Aug 13 04:58:57 wel kernel: sd 2:0:0:69: [sdei] tag#5 FAILED Result: hostbyte=DID_ERROR driverbyte=DRIVER_OK

Aug 13 04:58:57 wel kernel: sd 2:0:0:69: [sdei] tag#5 CDB: Read(10) 28 00 08 1a b8 00 00 04 00 00

Aug 13 04:58:57 wel kernel: blk_update_request: I/O error, dev sdei, sector 135968768

Aug 13 04:58:57 wel kernel: device-mapper: multipath: Failing path 128:160.

Aug 13 04:58:57 wel multipathd: sdei: mark as failed

Aug 13 04:58:57 wel multipathd: EFPdata066: remaining active paths: 1

Aug 13 04:58:58 wel kernel: sd 2:0:0:69: [sdei] tag#1 FAILED Result: hostbyte=DID_ERROR driverbyte=DRIVER_OK

Aug 13 04:58:58 wel kernel: sd 2:0:0:69: [sdei] tag#1 CDB: Read(10) 28 00 04 fb b4 00 00 01 00 00

Aug 13 04:58:58 wel kernel: blk_update_request: I/O error, dev sdei, sector 83604480

Aug 13 04:58:59 wel kernel: sd 2:0:0:59: [sddp] tag#1 FAILED Result: hostbyte=DID_ERROR driverbyte=DRIVER_OK

Aug 13 04:58:59 wel kernel: sd 2:0:0:59: [sddp] tag#1 CDB: Read(10) 28 00 08 1a 7c 00 00 04 00 00

Aug 13 04:58:59 wel kernel: blk_update_request: I/O error, dev sddp, sector 135953408

Aug 13 04:58:59 wel kernel: sd 2:0:0:59: [sddp] tag#2 FAILED Result: hostbyte=DID_ERROR driverbyte=DRIVER_OK

Aug 13 04:58:59 wel kernel: sd 2:0:0:59: [sddp] tag#2 CDB: Read(10) 28 00 08 1a 78 00 00 04 00 00

Aug 13 04:58:59 wel kernel: blk_update_request: I/O error, dev sddp, sector 135952384

Aug 13 04:58:59 wel kernel: device-mapper: multipath: Failing path 71:112.

Aug 13 04:58:59 wel multipathd: sddp: mark as failed

.........

Aug 13 05:33:05 wel kernel: sd 2:0:0:41: rejecting I/O to offline device

Aug 13 05:33:05 wel kernel: device-mapper: multipath: Failing path 69:64.

Aug 13 05:33:06 wel kernel: sd 2:0:0:25: rejecting I/O to offline device

Aug 13 05:33:06 wel kernel: device-mapper: multipath: Failing path 67:64.

Aug 13 05:33:06 wel multipathd: sdba: mark as failed

Aug 13 05:33:06 wel multipathd: EFPdata022: remaining active paths: 1

Aug 13 05:33:06 wel kernel: sd 2:0:0:46: rejecting I/O to offline device

Aug 13 05:33:06 wel kernel: sd 2:0:0:46: rejecting I/O to offline device

Aug 13 05:33:06 wel kernel: device-mapper: multipath: Failing path 69:224.

Aug 13 05:33:06 wel multipathd: sdcq: mark as failed

Aug 13 05:33:06 wel multipathd: EFPdata043: remaining active paths: 1

Aug 13 05:33:06 wel multipathd: sddf: mark as failed

=================================

 

I want to RedHat

- Shutdown failed path at the very first time of any path error.

- Not recover path automatically.

- And cluster should not move single path node.

 

what parameter should I modify on "multipath.conf"

I tested.

- "1 queue_if_no_path"  -> "0 queue_if_no_path"

- "no_path_retry fail"

but failed, multipath did not work, cluster moved node.

 

Do you know any other parameter and values that can meet this condition?

 

thanks for help.

Outcomes