SQL and Fault Tolerance

[rpreston]

I'm looking at providing a fault tolerant SQL solution, without having to use clustering software.

Forgive my lack of SQL knowledge, I'm a hardware guy and I'm trying to provide a solution based on my limited SQL experience, so I may be asking some stupid questions here.

So, would it be possible for me to have a SQL array configured on a NAS box, which is duplicated across the network to a SQL array running on a second box using DFS/FRS ?? If not, is there any way in which you could have the two arrays duplicated across network links ??

I'm keen to avoid clustering as we've had problems in the past with it. We'd like the two NAS boxes to be independent of each other.

Any help or advice would be greatly appreciated.l

Thanks.



Rob

 

[itsp1965]

From what I have read, NAS boxes are designed really for serving files and not block storage which a database requires. Although some companies purport to it being effective solution there is no clear cut answer regarding this. Have you thought about setting up another array and mirroring it to the existing one (RAID 10 or 50 don't remember which one).

 

[rpreston]

itsp1965,

Thanks for your reply.

Mirroring is exactly what I'm trying to achieve between the arrays, but I want to do this across the network rather than on two arrays physically attached via SCSI to the same box.

We had a problem with clustering recently where one of the two boxes connected to the external array had a fault. The resultant state that the OS was left in meant that we couldn't work on the faulty box because we couldn't be sure it wasn't going to bring down the box that was still working and, if that was the case, if a simple reboot would then have given us access back to the external array.

Because of this, we're now trying to look for a solution that gives us two copies of our SQL array, while also ensuring that they are not in any way tied together. The only way to do this is to mirror the two across the network, but I'm not sure what I can use to do that as I'm not sure what the potential impact of the data being SQL databases is. If it was a simple file structure, I wouldn't be concerned, but I don't know if SQL needs something specific to be mirrored ?

Hope that makes sense.



Rob

 

[itsp1965]

Instead of going the NAS route have you thought about implementing a SAN (I know it is really expensive) otherwise you can have two servers perform SQL replication. I don't really know the pros and cons of this, but hopefully one of the practicing DBAs on this forum can chime in.

 

[rpreston]

A SAN would get round all of our problems but cost, unfortunately, prevents us from going down that route at the moment.