Solved: FlexGroup configurations

Tas · ‎2019-03-15

I've seen the FG Best Practices guide and the manual.

Both seem to imply that disks, RG-types, and RG-sizes should be the same across the board when creating FG's.

Is this a best practice for best performance?

1. What if I have an ADP RG, which will have a fewer number of drives in the first RG? Is this okay?

2. The next four or five RG's in the aggregate will have our standard RT-type, 24-disc/4TB's. Would this be okay

The next node's aggregate may be all 24-disc RT-type, 4TB, 6TB or 8TB. Will this be allright?

Thank you

TasP

CHRISMAKI · ‎2019-03-18

Hi Tas,

The reason for this consistency requirement, which I believe is a soft-requirement, is for consitant performance across constituent volumes. When a FlexGroup is created, it provisions a multiple of constituent volumes across the aggegates it is allowed to consume from. If half of the FG is on NL-SAS and the other half is on SSD, access performance will vary wildly, depending on which constituent volume a file got written.

Since it appears all of your aggregates are going to be NL-SAS, I wouldn't be overly concerned about the differences between them, you likely won't notice much of a difference, if any.

View solution in original post

CHRISMAKI · ‎2019-03-18

Hi Tas,

The reason for this consistency requirement, which I believe is a soft-requirement, is for consitant performance across constituent volumes. When a FlexGroup is created, it provisions a multiple of constituent volumes across the aggegates it is allowed to consume from. If half of the FG is on NL-SAS and the other half is on SSD, access performance will vary wildly, depending on which constituent volume a file got written.

Since it appears all of your aggregates are going to be NL-SAS, I wouldn't be overly concerned about the differences between them, you likely won't notice much of a difference, if any.

Tas · ‎2019-03-18

Thanks @CHRISMAKI . @tmac said the same. Now I have confirmation from two sources.