Hi there

 

We are in the process of quoting a tech-refresh of our two node Stretched MetroCluster (A8080 nodes, one at each DC)

We were initially looking at a Ethernet based MetroCluster replacement, yet because of Cisco’s horrific back lock, we are more or less forced to choose another Stretched MetroCluster at the replacement.

The question is if there is a recommended upgrade path for this?

In a standard HA setup we normally just add the new two nodes into the existing cluster, and move volumes and LIFs… pretty easy and with no down time.

 

But it is a bit different with a two node MC mainly because it is essentially two single node clusters.

If we were to add another two nods into the existing cluster, we would also require four FC switches in order to link up the FCVI cards which is currently linked up back-to-back between the two DCs.

 

So I see two possible ways…

 

  1. Headswap… from A8080 to A400 (should be possible with any down time?  Just switchover, replace node, recable, reassign disks, switchback, then add new shelfs, and move volumes…) ?
  2. Parallel setup and move workloads from one to the other… (most is vmware, so not that big a deal… but there are also some critical iSCSI LUNs which will require more work)

 

One other little possible issue is that the old cluster has a 6G Disk Loop and a 12G Disk Loop… both use the QSFP/MiniSASHD to Optical… but the old 6G loop cannot work with the A400 because to my knowledge it does not have any supported PCIe card that has QSFP connectors?  (not according to HWU).

But we do have four 6500N ATTO Bridges that we can use for this… so right now the plan would be the following:

 

  1. Introduce the ATTO 6500N bridges into the existing solution (there are plenty of fibers across the two DCs)
  2. Switchover node 1 to node 2
  3. Headswap node 1 to A400, and recable.
  4. Reassign old node 1 disks to new node 1
  5. Switchback node 1
  6. (same procedure on node 2)
  7. Attach new disks to both nodes.
  8. Create new root aggr0 on new disks for both nodes (move configuration to new aggr0… (described somewhere…)
  9. Move volumes to new disks.
  10. Remove old empty aggregates and unassign
  11. Remove physical disks

 

 

Any other suggestions?  And have anyone else tried this?

I think I read somewhere that it was actually not possible to expand a two node MC to four nodes… is that really correct?

 

/Heino