Boot from SAN, why? what is the recommendations?
What is the recommendation for doing boot from SAN?
I have /,/var,swap partitions, with of them is recommended to put on the SAN for sure?
Generally, what are the advantages and the disadvantages for using boot from SAN?
As I see it, the big advantage to create /,/var on the SAN because it is good for DRP/DRO for copy those's luns on a backup site, which have the same server(hardware).
I checked also the performance, and it seems that the i/o performance was improved after I moved to boot from SAN.
NOTE: my SAN is McData4700 and AMS-500(HDS storage with 2GB cach)
[607 byte] By [
shayberya] at [2007-11-27 3:18:54]
# 1
One of my esteemed colleagues has argued very strongly that booting cluster nodes from SAN is a bad idea simply because of the complexity it added. Complexity does seem to be the enemy of availability. So, having tried to set it up and seen some problems from time to time in the lab, I'm inclined to agree with him.
Not that this doesn't work though. I've worked with other customers who wanted to boot from SAN and have done so successfully.
Tim
# 2
Which kind of complexity you are talking about?
It seems that it is realy easy to do , cluster booting from SAN.
Can you give me examples of problem that you could have by using Cluster booting from SAN?
What about the swap? where do you think is better to locate it SAN\localy?
# 3
Compare a root disk on a locally attached JBOD disk with one attached via the SAN:
JBOD: server -> HBA -> JBOD physical disk
SAN: server -> HBA -> switch -> zoning -> array -> vLUN masking -> vLUN to physical disk
For switch, you have firmware and configuration
For lun masking you have firmware and configuration
for vLUN mapping you have configuration
Furthermore, the real disk might not be colocated with the server so you cannot easily tell when you are next to one, what the state of the other might be. JBODs are usually with <20 metres of the server.
In short, there is much more that can go wrong with storage on a SAN that with a simple locally attached. Break any bit and you might not be able to boot one or both nodes. Of course, you can avoid some issues with redundant SANs, but then if you have one SAN admin who makes one fatal mistake twice...
That's not to say they don't work, but I'd definitely like fewer things to worry about if I had to manage such a configuration.
Of course, the down side of this is that you lose mobility. Here there might be something to be said from booting via NFS. [ discuss ;-) ]
Tim
