mirroring plexes across arrays
Are there any cons to mirroring plexes for a single volume across two different arrays for redundancy? I'm familar with the process of mirroring plexes and use it often to migrate a volume from one lun to another lun but was unsure what was the longterm effect of mirroring across arrays.
Mirroring across arrays is fully supported and expected...
In terms of cons:
1. Writes to have to go to both arrays. If one array is slower than the other, write speed is determined by the slower array.
2. It is a bit more suspectible to human misconfiguration. For example, let's say you set the plexes up, 1 on each array. Then you need to grow the volume, and you are on vacation, and some junior guy does it in the middle of the night, and he grows the volumes -- but the subdisks for both plexes are coming from the same array. You've lost your redundancy in case the other array fails.
Two mitigation choices for the second issue:
1. You can check for that using the SORT Risk Assessment: https://sort.symantec.com/home (see Custom Reports).
2. Recent versions of VxVM have some rules you can setup so that volume grows use the same arrays as the original plexes.
