Veeam & SRM

[jessethomas]

Hi Everyone

We are in the midst of a compute and storage refresh. In addition to new hardware, we are moving to NetApp storage (from EqualLogic), a more balanced two-site active/active setup and adding in SRM to (hopefully) improve our DR posture — for about 175 guests.

We have been using Veeam Replication for about 3 years now, to protect our most important machines, but without a formalized "action plan" for recovery in the event of a complete site loss. Veeam has served our needs very well, but we have been looking for something more automated in the event of a true disaster (hence SRM).

We are new to NetApp and SRM, but have been learning furiously and working with some partners on the setup over the last few weeks. We recently learned that SRM (with SAN snapshots) doesn't provide application-consistent recovery and that there really isn't an (easy) method for granular failover (e.g., failover a single failed VM). These are features we have grown to appreciate and rely on in Veeam.

To keep things simple and in an effort to use our storage space efficiently, we don't want to have Veeam replicas AND SRM replicas. One idea we've had is to use Veeam to "power" the replication between sites and leverage the granular recovery options, and then integrate SRM for failover automation, but are not sure if this is possible.

We are engaging our Veeam SE, NetApp SE and others on this, but I'm also hoping someone here might have some experience in this area, and would be willing to share their experience as well (so we can get as complete of a picture of the options as possible).

I am aware of the forthcoming Veeam Orchestrator and think this would likely be an ideal fit for us, but there is also a desire to address the DR automation as part of the refresh.

[skrause]

So when you talk about automation, do you mean orchestrating the recovery tasks or "system detects a fault and automatically turns everything on?"

Veeam, with failover plans, can do the orchestration of failover very well once you configure it. But it was designed to NOT do the second style on purpose. Requiring some level of user input to "press the red button" is the core reasoning behind that. The number of false or misleading variables that could lead a fully autonomous system to failover is very high and planning for all of them is virtually impossible.

Speaking from (current) experience, failing over in a DR scenario sucks and if you can avoid un-necessary failovers your work/life balance will be better.

[jessethomas]

Steve - we are only talking about "orchestrating the recovery tasks". We have no interest in any type of automatic solution, for the reasons you indicate.

We are somewhat familiar with the Failover Plans in Veeam. Any comments on how they compare to the functionality in SRM?

[skrause]

We ditched our fledgling SRM implementation when we bought Veeam for a couple of reasons:

Granularity: We like being able to choose by VM and not by datastore.
Cost: SRM licenses were expensive plus the added cost of adding replication to our new storage arrays was prohibitive.

The SRM implementation we had pre-dated my hiring and we never really tested it so I don't have any operational experience with the differences.