2011-06-14 05:53 PM
I asked a NetApp tech, they said it's impossible, but I want a second opinion.
We currently have two identical single headed (single controller) FAS2020. They are connected to a Qlogic SanBox Fibre switch. Currently they are setup as stand-alone filer with SnapMirror enable for our DR solution. They are placed in different physical locations (200m apart).
We would like to setup them up as cluster so that one will server data while the other one failed. Meanwhile we need to continually keep them as our DR solution. I assume SyncMirror is what I should be interested in?
We need the data on filer to be mirrored at all time even after a take-over and give-back. They are to be setup as LUNs.
I believe the configuration I should implement is Mirrored Active/Active?
I understand that additional FC-VI adapter must be installed on both filers as cluster interconnection. Do I need these interfaces installed for Mirror A/A mode or the InfiniBand (IB) cluster adapter?
Solved! SEE THE SOLUTION
2011-06-14 10:22 PM
Hi, the NetApp Tech was correct. The 2020 has no PCI slot to install the necessary Interface Cards to implement the HA (Syncmirror) Functionality you want to get.
This is only possible with the 3000 and 6000 Series, because one also needs additional path to the diskshelfs (which is not possible with the 2000 Series as well).
So you either stay with your SnapMirror Solution, or you can convert the the two "single" 2020 Controllers on each location to HA Controllers and provide at least in the local DC the HA functionality. With DR Failover based on SnapMirror.
Hope this helps,
2011-06-15 12:32 AM
I think as both the tech and another poster has said, not something you can do with a 2020, as it has no expansion slots for the FC-VI adapter and no support for metrocluster…
Are you trying to get an instant automated failover between the two controllers?, as even with metrocluster, you have the same NetApp rules, which is that one controller is the owner of the disks, (so delivers LUNS, Shares, etc) while the other controller takes them over in the event of failure, both controllers are active, so can both deliver LUNS, Shares etc, but not the same LUNS at the same time, same as with any other filers, including a 2020HA.
If you want to automate the process between your two 2020’s maybe worth looking at Protection manager as this can help you to automate that can of failover, failback scenario, you will also need to look at the apps and data you are serving though, to see how they would cope with an interruption and break such as that…
Hope that helps, even if its not the answer you wanted, you need something from a 3210 upwards for metrocluster support (don’t think you can do it with a 2040)
2011-06-15 01:38 AM
Firstly, syncmirror is not a cluster failover solution so it does not sound like what you want. Syncmirror is essentially adding another layer of RAID; it mirrors the content of an aggregate onto a second set ("plex") of disks. Syncmirror is there to save you if you believe that RAID-DP isn't enough. It will not save you from a controller or chassis failure.
Snapmirror over Fibre Channel is right out, because you can't add the expansion card on a 2020. I doubt that Metrocluster can be done on a 2020 either given that it is designed to have two controllers internally.
The next best thing available is to use vfiler dr with SnapMirror sync/async. The failover won't be automatic, unless Protection Manager can fail over the vFilers for you in the event of an outage.