Recovery after a total disaster of the cluster.

[JBC]

Hi all

For testing purpose, I try to recover a Fas2520 cluster mode 8.3.2 from almost factory configuration with external copy of system configuration backup.

I didn't deal with users data, I just want to recover node, aggregates, volumes, qtree, svm and so on configurations.

I made several tests but I think something missing in my procedure. Each time I got ghosts aggregate: newly created root aggregates at boot after zeroing disk are only in WAFL and all recovered aggregates from backup file (included old root aggregates aggr0) are only in VLDB and don't got any attributes (no size defined, not shown in "aggr show" command)

How i proceed:

1 -    Initialize FAS : disable the HA capability, disable storage failover for both nodes, Halt the system on each nodes and boot in maintenance mode, Make all aggregates offline and destroy it, Remove partitions and disk ownership, Boot again with option 4 and initialize all disk (zeroing).

2 -    Boot each node from Service Processor  and put basic configuration (IP netmask gateway ...)

3 -    On each node, enable "network options switchless-cluster modify " , rename node with previous name, import external copy of backup file, and restore the node from this backup file.

4 -    On node A , restore cluster with "configuration recovery cluster recreate -from backup " then join the node B to cluster with "system configuration recovery cluster rejoin -node xxx

5 -    Finish process with "system configuration recovery cluster modify -recovery-status complete" and  "cluster ha modify -configured true" and "storage failover modify -node xxx -enabled true" for each node

Sorry if my english is not correct and thanks in advance for reply.

[aborzenkov]

I doubt very much that what you expect is possible. Configuration backup contains cluster configuration (content of replicated databases). It can be used to restore access to existing data after those databases are corrupted beyond repair. But recreating all aggregates during configuration restore would be really disastrous for this use case. May be there are some diag level options, but it is definitely not something I expect to happen by default, normal usage.

Root aggregates are only in WAFL because they obviously have different UUIDs so are not known to configuration you restored.

What you do looks more like replicating cluster configuration to different cluster than to recover existing cluster. As far as I know there is no tool to do it. I would rather look at using ZAPI to capture configuration items you are interested in and replaying them on new cluster. In principle it can even be done using standard CLI

[JBC]

Ok, thanks for reply. I will have a look on ZAPI  and try this way