MPIO preferred or weighted paths not persistent through reboot

[transparent]

Hello all,

I'm having an issue where our MPIO paths are not 'sticking'. We have a 3-node Virtual SAN/Hyper-V cluster running on Windows 2012 R2 and Starwind Virtual SAN v8 (latest build as of today). Everything works 100% by the way - this is just a failover/failback issue, so I don't *think* its a Startwind issue per se, but thought I'd start here.

My ideal setup would be to have 'Failover only' MPIO policy, and I've set the preferred path to the localhost target and the path to the other two hosts as 'Standby'. When I go into Disk Manager and go into the MPIO properties of the disk, I have the option to check off "Preferred path" check box, but after clicking on 'OK' and going back in to look, the box has been cleared. This is on all three servers by the way.

Next I decided to try Weighted Path instead and give localhost a weight of 0, and the others as 100. That seems to work and I can see all iSCSI going over the loopback. If I disconnect the localhost target, I can see iSCSI traffic flowing over to the other two hosts without skipping a beat - great. However, if I reboot the server all of the weights revert to '0' and its effectively just a round-robin at that point. I have to manually go in and apply those weights again. Doing cluster updates will no doubt affect performance as the hosts reboot.

Anybody know what I'm doing wrong, or if there is a work around for this? Its my last stumbling block before going production.

I've searched the forum for a similar issues and found an example here, but sadly no resolution, the affected person just left it as round robin, but I'm not willing to give up on the local performance of the loopback adapter.

Thanks in advance...

[darklight]

Hi mate!

Have you tried to go to iSCSI initiator settings and point to properties and device options there? There is an MPIO policy selection box, so you can define whether round robin or failover only is used for each device.

[transparent]

Thanks for the suggestion. Yes, I've double checked the settings both in the iSCSI Imitator (this is where I've been setting the weight of each path that don't survive a reboot), as well as through the disk properties in Disk Manager (where I've tried to set the preferred path, but the check box gets cleared after hitting OK and reopening).

[Slingshotz]

I have a similar issue that I think is related to yours and I'm also running a 3 node setup. I have set the iSCSI initiator MPIO device details to be failover only and pointing to the local adapter 127.0.0.1 as the target portal and active as per Starwind's instructions. When the server gets rebooted, the MPIO always changes to one of the other servers to be active target portal. It never reverts back to the local adapter unless I go back and manually make the local adapter active again. I have never noticed any issue with having the active path on either the local adapter or one of the other servers, but I've always been a bit concerned since the instructions clearly state to make the local adapter as the active target portal.

It is probably happening to you as well and as a result it resets your preferences when it changes your local iSCSI path state to standby. Just my guess but I have never tried using that option in the MPIO properties.

[transparent]

Yeah, it sure sounds like the same issue. At the end of the day it 'works' and fails over properly, but I've noticed that I get much better performance when a host is using its local storage using the loop back acceleration. At this point I'm wondering if there is a script I can setup to run after a reboot completes to fail back properly. I've seen others report this (not just with Starwind), but no verified answer as to why this happens. I'm not near the cluster right now, but I'm thinking that the path id's are changing after reboot, so it treats it as new paths with default weights. This does not explain the 'preferred path' issue - as that setting doesn't hold past clicking 'OK' on the settings panel.

Perhaps related, but I've noticed a DSM Settings button, and one of the check boxes is 'Enable path verify'. I can't seem to find what's that for, but I've seen suggestions to enable that, but I'm wary of changing settings that I'm not familiar with and possible adverse affects.

Thanks for your input, much appreciated.

[darklight]

I believe it is possible to create a startup PS script that will do a proper configuration of MPIO settings:
https://technet.microsoft.com/en-us/lib ... .630).aspx
Will try to do this by the weekend.

[Tarass (Staff)]

Darklight, thank you very much for you contribution on the case.

2 All Take a look at mpclaim /? command.