I am wanting to implement Flexgroups to house user shares (due to a number of the wonderful things a Flexgroup would do for me).
The pushback I am hearing, (yet cannot find any evidence of the veracity of these claims) is:
1. That data on a constituent volume becomes unavailable should the owning node fail (the constituents do not fail over). This does not sound logical.
2. If I enforce quotas (when 9.5 becomes GA), I'm enforcing a quota on a single constituent volume, and still need to worry about constituent volumes filling up, thus Flexgroups do little to help in terms of balancing capacity.
Can someone please provide an answer as to whether any of this is correct?
Like you, I could find no documentation to back up these supposed authoritative claims. I will include node failure in the test plan of my design. When the 9.5 simulator becomes available, I can test quota enforcement as well.
"Constituent and Member Constituent and member are interchangeable terms when referring to FlexGroup. These items are the underlying FlexVol volumes that make up a FlexGroup volume and provide the capacity and performance gains that are seen only with a FlexGroup volume."
And check out page 49 of the TR for 9.5 quota enforcement.