Cisco Hyper Flex Replication Factor

[AbcKdc31268]

Hi All, 

I am new to Cisco HyperFlex. We currently have a 3 node Cisco Hyper Flex cluster. Each node has 7TB*6 disk per node i.e. 42 TB per node. We currently have Replication Factor of 3 with the useable capacity of 38 TB. We are now planning to add a new node with same configuration and storage. With Replication factor of 3, as per my knowledge data is replicated is 3 times. So, we can have a 2 node failure (not at the same time but one after another). So if we add a new node with same storage configuration I expect my usable capacity also to grow i.e. around 75 TB. Is my assumption correct or am I missing something.

Thank You.

[RedNectar · Accepted solution]

Hi @AbcKdc31268 ,

The best place to answer this question is to go to the EXTREMELY HARD TO FIND https://hyperflexsizer.cloudapps.cisco.com/

(Google just doesn't seen to know about it)

I did a quick "customised" config for you configuration with the default assumptions about compression and deduplication and got...

So that's the sizing question answered. I encourage you to do your own sizing at https://hyperflexsizer.cloudapps.cisco.com/

But you also said

we can have a 2 node failure (not at the same time but one after another).

Well - that's true... BUT ...

1. It really doesn't matter if it is at the same time or one after the other. If you loose 2/4 nodes you are left with all your data but only two copies - possibly one one copy of some of it until the system re-balances, so I guess the "one after the other" makes sense
2. The chances of loosing two nodes is pretty slim, but if you think it is a possibility, then you should be planning on 5 nodes, not 4, that way you can maintain RF3 even with 2 nodes down (which is why Cisco recommend 5 nodes with RF3 - although the salespeople trying to make a sale probably won't mention this)

And one more thing - probably too late now - if you have ALL FLASH systems, RF2 is probably good enough, because if you have 4 nodes, and you loose one, the system can rebuild that missing data VERY QUICKLY (I hesitate to quote times, but the thing is, all 3 nodes will work together to rebuild the missing data after the timeout for node-down(= 2hrs) has expired). And now you are back to RF2 again only after a few hours. And you still can loose another node and go through it all gain if there is headroom capacity.

However, it seems you are on RF3, so that's not an option, but I've included it in the reply for the benefit of anyone else reading this in the future.

[AbcKdc31268]

Hello @RedNectar Thanks a lot for the reply and again thanks for the running the sizer for me. It seems the storage capacity will increase in our case. As we have already gone with RF 3, so, as per your saying we can have 2 node failure at the same time and the cluster will still function it will not go into read only mode. If this is true we are fine with the expanding the cluster as we might even expand the cluster with more than 5 nodes in future.

Thanks a lot again.