Unsolved
This post is more than 5 years old
1 Rookie
•
62 Posts
0
2410
VPLEX MPIO algorithm for Aix 7.1
What is the recommended native MPIO path failover setting for vplex local geo sync 5.4 and AIX 7.1. We are currently using algorithm "fail_over" . This does't seem to be using all paths to vplex front end ports. I am planning to use Round-Robin. Does anyone have more experience with this or any suggestions? I have checked host connectivity guide and vplex host multipathing documents. I could't find a recommendation for MPIO with AIX hosts. I know that powerpath needs to be set as "Adaptive "
echolaughmk
522 Posts
1
October 22nd, 2015 07:00
Round_Robin is the one most widely used in an Active/Active setup like VPLEX from my experience. The fail_over algorithm will only use a preferred path for the most part and failover to a standby path when that preferred path fails. Using Round Robin should get you the results you desire where all active paths are utilized for I/O requests. I don't believe there is a "recommended" policy from EMC's perspective since the software is IBM's.....from their perspective they will all work and it is just a question of how you want MPIO to work where you chose the different algorithms and policies. I have used RR before with AIX and VPLEX and haven't come across any issues for the most part.
HTH
Ayrton_Senna1
1 Rookie
1 Rookie
•
62 Posts
0
October 23rd, 2015 19:00
echolaughmk Round-Robin seems to be the way to go since VPLEX is active/Active and also helps to ride through Code upgrades on VPLEX when one side of Directors completely reboot. However, in some cases of RoundRobin I have seen the Host continue to use the paths that are partially failed due to bad cable. Have you encountered this scenario with VPLEX ? How robust is MPIO in detecting faulty paths ?
echolaughmk
522 Posts
0
October 26th, 2015 04:00
Hi,
Yup...I have seen that with MPIO in general....both with bad cables and even if there are SFP issues on the switch. AIX can see them as aborted commands and will just retry and throw SCSI errors in an errpt versus fail the path in the MPIO config. PowerPath might curtail some of this because if has some additional error detection on the path, but I can't say for sure that it will catch all of them either. From what I have seen is that if there is a physical path failure, MPIO will catch it and re-route (cable pull, HBA failure (no light), zoning removed accidentally, etc). For some of the logical path failures, I would suggest they are tested as they might not all spark a path failure incident as desired (CRC errors on the SFP, dirty cables, etc). There might be tweaks in MPIO where you can set some thresholds to invoke a path failure but not all of them are set by default with the default EMC ODM's from what I have seen. IBM might be able to help suggest some tweaks in that area as well.
HTH