1300 DRS migrations across 2 clusters in 24 hours

[marshallmathers]

im reviewing some vcenter logs and found a site with 1300 DRS migrations across 12 clusters 775 are from 1 of 12.  500 are from a second cluster .

so basically across 2 cluster im seeing almost 1300 DRS migrations in 24 hours !!!

these 2 clusters are both located in the same Data center / Physical location.

is this normal? seems awfully high ? What can i do to investigate ? Where to start ?

looking for guidance .

thank you .

[jburen]

First look at the DRS configuration of both clusters. Maybe it's too aggressive...

[marshallmathers]

the DRS settings are either manual, partially automated or fully automated with a migration threshold of either 1,2,3,4,5

are you saying if its fully automated and a migration threshold of 5 to lower it to a 4?

would there be any other settings i should be checking ?

[sjesse]

Unfortunetly its not just one setting, so any suggestions from us would be a shot in the dark, I really suggest  reading the documentation especially the below link. Take a look at the different settings you can set and figure out the best combination. SDRS moves based on io and space and each of those have different settings. If you have I/O imbalance threshold set to 5, that's at the very end of the aggressive side the slider, its probably better to be set in the middle, and then increase until servers are moved when there is a need and not every slight imbalance. I manage a few different environments and each have different settings based on the needs of that environment. Don't over look the space settings, if you have very equal full datastores and the space utilization difference is very low it may move vms every time they grow and that threshold is crossed.

Setting the Aggressiveness Level for Storage DRS

[marshallmathers]

the clusters i am seeing this on are not storage clusters they are host clusters. 

[sjesse]

My apologizes not sure why I read that as SDRS, then yes I'd lower it to 4, then if that seems like too much then lower to again. I'd look at the history though and see if there certain vms that are moving more then the other ones, they might have odd workloads that drs can't handle, and you may need to either pin those vms to a host using drs affinity rules to prevent them from moving so much.

[marshallmathers]

thanks i will take a look .