#SVM-DR - when are vserverdr-snapshots removed?

1 messages · Page 1 of 1 (latest)

foggy spade
#

We have an SVM-DR relationship where a destination aggregate (one of several) where a couple of volumes were stored, filled up. We've since cleaned up, so the destination has plenty of space, and snapmirror has resumed (as seen by "snapmirror show" and network traffic).
I can see the snapshots arriving on the destination cluster, but when are they removed? The two volumes, vol1 and vol2, now have about 100 snapshots each on the source side. On the destination side, vol1 has 100 snapshots, and vol2 is at about 40, increasing as they are replicated across.

I guess my question is - are no snapshots released (from vol1) on the source side until all volumes in the SVM-DR is replicated across?

quartz spindle
#

Yikes, I may need to "phone a friend" on this one but I believe yes that is the case. We need the full SVM to replicate before we release the snapshots. That gives you a recovery point that is common across the volumes in the SVM so you can recover the SVM to that common point. This first replication will mean some of those snaps linger for a while but after that they should only be around for a few min as the SVM updates on each SM pass.

foggy spade
#

This isn't the first time I've seen this happen, and I think I recognize the pattern from last time. It makes sense, too, so I'm fine with it. However, it does bring us back to another topic which I would love to be adressed: Some sort of progress status for snapmirror transfers. X out of Y TBs transferred, X out of Y snapshots replicated (even if Y changes during the transfer, just seeing the progress of X helps).

#

Today I'm just watching the network traffic on the replication interfaces to see if anything is happening - "snapmirror show" has shown no progress for the last couple of days.

#

...and the WebGUI/sysmgr show "finalizing" as status

fierce cedar
#

Snapmirror is very different with XDP as compared to DP.

A baseline snapshot is created and the base volume is transferred. If you do a snapmirror show frequently enough you will see it transferring then finalizing for each snapshot. Effectively, the base is transferred and committed and then each snapshot is transferred and committed until it catches up.

At that point, if I recall, there will bend up being 2 snapmirror snapshots for moving forward.

foggy spade
#

Yeah, I can see the two snapshots (which appear to be named the same on all volumes part of the SVM-DR relationship).

#

I'm still not sure how to figure out what the snapmirror process is currently doing - as far as I can see I have to manually go through all volumes part of the SVM-DR and do a "snapmirror show -destination-path svm-dest:vol1", "... :vol2" etc until I've been through all volumes and seen which have a status of !Idle? (and then look at Transfer Snapshot etc)

#

I'm on 9.7, if that matters.

fierce cedar
#

There is an -expand option you can use. Here is an example

set diag
snapmirror show -expand -vserver dest-svm -fields source-path ,type ,destination-path ,state,status ,total-progress,transfer-snapshot,healthy -status !Idle -sort-by source-path

#

Note: in order to sort, you need to be in diag mode

foggy spade
#

This is awesome, thank you. For some reason I've missed the -expand option.

fierce cedar
#

Take advantage of the sorting too. With lots of mirrors, the sorting can be very handy

foggy spade
#

Yeah, I saw - but what's the rationale behind requiring diag-mode for that? That seems... excessive?

fierce cedar
#

That’s a question for Netapp. It’s so useful. It should be in admin mode

atomic mango