Destruction Testing

John Paget Bourke
Occasional Advisor

Destruction Testing


I have a small cluster 2x DL380 G5, 2x SAN switches, 1x MSA 2312, RHEL 5.3 and multipath.

I want to do some Destruction Testing if this configuration so that I know it survives failures.

All this redundancy is great, but how do I see it working.

So how do I simulate various failures such as

1. Disk failure
2. PSU failure
3. Array controllor failure
4. HBA failure
5. Switch failure
6. Cable failure (unplugged)
7. Multipath software failure

(I asked 6 because I do not know if you can jst pull a fc cable and insert it again, you cannot do that with SCSI).

Any thoughts ?

Mark Wibaux
Trusted Contributor

Re: Destruction Testing

1. Pull a disk while the system is running. They're hot swap so no issue there.

2. Pull a power cord from one of the power supplies

3. Pull one of the controller modules. They are hot swapable.

4 & 6. Pull the fc cable from the HBA. This will fail that path from the host and show if the other path is working fine.

5. Pull the power on one of the switches.

7. Not sure what you are trying to test here. From what I recall the multipath functionality is built in to the FC HBA drivers in linux. If the driver fails then I would expect that host to stop seeing the SAN altogether.