change path selection policy from fixed path to round robin

[xbradshr]

two questions

1. is it safe to change the path selection policy from fixed to round robin with servers on the datastore(round robin does work in our environment)

2. Can "lost access to volume" messages be caused by too many servers using the same path to datastores. A lot of the servers here are using a couple of paths on fixed path , instead of all four paths on round robin. I was wondering if the lost access to volume messages were caused due to heavy traffic causing the "heartbeat" data writes to time out.

[rcporto · Accepted solution]

1. is it safe to change the path selection policy from fixed to round robin with servers on the datastore(round robin does work in our environment)

You can, but first make sure your array vendor supports this configuration.

2. Can "lost access to volume" messages be caused by too many servers using the same path to datastores. A lot of the servers here are using a couple of paths on fixed path , instead of all four paths on round robin. I was wondering if the lost access to volume messages were caused due to heavy traffic causing the "heartbeat" data writes to time out.

Check if helps: VMware KB: Host Connectivity Degraded in ESX/ESXi

[Kauy]

In addition, verify if your Guest OS configuration/topology also supports RR. There are some specific recommendations in case you have MS Clusters running on VMware (VMware KB: Microsoft Clustering on VMware vSphere: Guidelines for supported configurations). I have also seen some issues when setting up Veritas clusters hosted by ESXi 5.5 on RR (Especially when PSP IOPS Limit is set to 1).

[impranayk]

1. is it safe to change the path selection policy from fixed to round robin with servers on the datastore(round robin does work in our environment)

          Yes, you can change the PSP from fixed to RR or RR to fix without any downtime. Make sure that you are changing it form GUI, as doing it from esxcli may revert back to old policy after reboot.

2.Can "lost access to volume" messages be caused by too many servers using the same path to datastores. A lot of the servers here are using a couple of paths on fixed path , instead of all four paths on round robin. I was wondering if the lost access to volume messages were caused due to heavy traffic causing the "heartbeat" data writes to time out.

          This may be due to ESXi Management traffic and datastore traffic going on the same subnet. Every 16 seconds the ESXi host sends a heart beat to the LUNs it knows about presented by an array . If the array doesn't respond then the host marks the LUN(s) as inactive and  this message " Lost access to the volume" is presented

VMware KB: Host Connectivity Degraded in ESX/ESXi