Network and Storage Protocols

are snapshots dedupe aware in 8.0.1?

RAJA_SUBRAMANIAN
3,185 Views

Hi All,

I have a new dual controller FAS2040 running ONTAP 8.0.1 7-mode.  The only use for this filer is CIFS for 300 users, where both dedupe and snapshots will be beneficial.

TR3505 mentions that snapshots are not dedupe aware (snapshot size balloons when dedupe is run) and recommends creation of snapshots only after a dedupe run is completed.  Does this limitation still apply to 8.0.1?

Snapshots are higher priority feature, and I want to follow the default weekly/daily/hourly snapshot schedule.  What would be the best dedupe + snapshot practice for my environment?

Further details about my setup:  FAS2040 with 24 x 450GB SAS disks. Each controller has 9D+2P+1S with single 3TB aggr0, vol0 (150GB) for OS and vol1 (2.7TB) for data.

Thanks in advance!

- Raja

1 ACCEPTED SOLUTION

Darkstar
3,185 Views

This is a WAFL limitation and it is still present in 8.0.1. Block pointers in Snapshots cannot be changed by the filer.

We always schedule our SIS jobs to run at 23:00, one hour before the nightly snapshot. This means that at least the nightly+weekly snapshots (which will be around for some time) are well deduplicated. The hourly snapshots will probably not be 100% deduplicated (i.e. changed data is not), but since they're not around for that long (a few days at most, normally) this is a tradeoff you can most probably live with.

In general you'll simply have to try it since it depends heavily on your data volume, change rate, and so on. But you won't feel any ill sideeffects from doing it like this

-Michael

View solution in original post

1 REPLY 1

Darkstar
3,186 Views

This is a WAFL limitation and it is still present in 8.0.1. Block pointers in Snapshots cannot be changed by the filer.

We always schedule our SIS jobs to run at 23:00, one hour before the nightly snapshot. This means that at least the nightly+weekly snapshots (which will be around for some time) are well deduplicated. The hourly snapshots will probably not be 100% deduplicated (i.e. changed data is not), but since they're not around for that long (a few days at most, normally) this is a tradeoff you can most probably live with.

In general you'll simply have to try it since it depends heavily on your data volume, change rate, and so on. But you won't feel any ill sideeffects from doing it like this

-Michael

Public