I have a volume accessed via CIFS with local (hourly/daily/weekly/monthly) snapshots, VSM and then SV. This is a "normal" dataset, not an application dataset like what snapcreator or the various snapdrives will create. VSM happens every 15 mins. The SV schedule in the protection policy is configured to run 1 hour after the primary snapshots are created (including offset for timezone differences) for each daily/weekly/monthly.
My question is about the snapshot that is used to update the SV on the final leg of the cascade. Since I plan to start deleting data on the primary side after the local snapshot is complete via a post script it is important that the SV happens from the snapshot and not the active filesystem. The post script will only trigger on monthly snaps, and since the monthly SV copies will be used for long term storage and compliance the data that was deleted in the post-script must be present in the SV snapshot.
How does protection manager pick which snapshot to use when updating a SV destination from a VSM source?
I've been doing a good amount of testing today in my lab environment and have found the following information that could be useful for someone else with the same setup.
My environment is similar to what is listed in TR-3487 figure 4.
To preserve the data for long term archive I must update the SV relationship from a named snapshot copy. Even with snapvault.snapshot_for_dr_backup correctly set I was unable to get the normal PM dataset to update the SV from the recently created snapshot, it was always taking from the snapmirror base snapshot. It would only work if the dataset was created via snapcreator. This setup will work well as snapcreator will call PM to start the backup, which will go all the way through the cascade and update SM and then SV from the recently created snapshot. The schedule on the protection policy was reconfigured so that the primary and backup have no schedule, and the mirror still executes every 15 mins.