#ONTAP Mediator vs. ClusterLion for Metrocluster

1 messages · Page 1 of 1 (latest)

elfin jasper
#

Hey hey,

we are currently re-evaluating our infrastructure and I came to the point of checking the new features of the ONTAP Mediator in comparision to the old Tiebreaker Software.
As I currently understand it, I see no features of the ClusterLion, that get not immediatly solved by a 3rd location and an ONTAP Mediator.

Is there any benefit to a ClusterLion over an ONTAP Mediator if we already have a 3rd location to host the Mediator?

Relevant links:
https://prolion.com/clusterlion/clusterlion-metrocluster/
https://docs.netapp.com/us-en/ontap/mediator/mediator-overview-concept.html

Thanks for your input 🙂

sly harbor
#

ClusterLion is the only solution that works when your network is down/impacted. For example if you don't have two completely separate, passive network links to your Mediator location

#

also, ClusterLion can physically disconnect power from the switched-over nodes so that you can be 100% sure it's powered off (no split brain)

elfin jasper
#

With Mediator, if the Cluster A looses connection to the Mediator AND to Cluster B, it powers down. So the second point gets solved, no?

main prairie
#

Are you sure about that? I don't think Cluster A will power down in that scenario. I think each cluster will stay online and continue providing its own workloads. And MAUSO will not be initiated since Mediator connection is lost and NVRAM mirroring is not in-sync anymore.

elfin jasper
#

My contact on NetApp confirmed this exact question.

sly harbor
#

If Cluster A loses its connection to Cluster B first, then NVRAM sync will not work anymore and no automatic switchover can occur because it would mean data loss. So both nodes keep running

#

I'm pretty sure the other way round (first losing access to the mediator and then to the other cluster) will also not result in an automatic switchover but I can't find it written down anywhere right now

#

TR-4705

elfin jasper
#

I double checked with my contact and seems like this is right and I missunderstood the previous statement.

#

So if a SIMULTANEOUS loss of Mediator and Cluster B happens, Cluster A shuts down. But if Mediator is lost and then Cluster B OR Cluster B is lost and then Mediator, then firstly no MAUSO but also no split brain.
So the advantage of Cluster Lion is, that it has a separate way of knowing the status of both clusters (as it is not depended on the network) and can initiate the MAUSO, guaranteeing (near) continous data access.

#

Thank you for the input @sly harbor and @main prairie, this cleared things up for me.