Question for the Experts: HA MS-CS and SQL Server

[milanbenes]

Afternoon All,

HA experts, as we all know from experience, no matter how much our blue chip clients invest in HA, from time to time software like clustered databases, SQL Server running over Microsoft Cluster Service can have problems where even though the node with the DB is running fine, because for some reason the other node of the cluster became unavailable and not in a clean MS-CS way, the Cluster software decides not to work and therefore no DB - even though that is precisely what it should be solving.

Ok question, if we have clustered SQL Server running on Microsoft Cluster Service and the Networking Team take away the Network for 30 minutes, when the Network comes back will MS-CS pick up from where it left off and continue running fine or will it be screwed up ?

Thanks,

MilanBenes in 'I should stick to the backup tapes' mode.

[ratewhore]

One for the technical forum moderator...

[milanbenes]

you useless fecking cant just answer the question



Milan.

[AtW]

If its any good (like Sybase was) then it should use transaction log to sync all databases, otherwise there is no point in such clustering.

[ratewhore]

you useless fecking cant just answer the question



Milan.

OK, screwed...

[milanbenes]

AtW,

I am talking about the Microsoft Cluster Service software not crashing when the network goes down.

Secondly, a clustered database is a clustered database singular, one database clustered between two servers with a quorum disk giving an attempt at a singular highly available DB if one server goes down running the db then the db is moved to the other server by MS-CS - in principle - however MS-CS does not always do what says on the box hence my question.

To conclude, AtW, if you are not sure what I am talking and therefore don't know the answer please continue writing your SKA and avoid filling this useful thread with crap.

Thanks,

Milan.

[Viktor]

Have a look at this:

http://www.microsoft.com/technet/pro...g01.mspx#EIGAG

It's always better to have a cross-over for the heartbeat network, or a dedicated switch. The nodes see each other through the heartbeat network, the master remains the one who owns the shared disks (active/passive config).
If the passive node goes away in cluster admin, evict the node and add again the server. Normally the nodes should pick-up from the quorum log, but you never know...

Hope every little advice helps,

Viktor

[AtW]

[milanbenes]

thanks Vic.

both nodes will lose their network.

Looks like MS-CS will be fecked.

Oh well one good thing about MS, it should work after a restart.

Ok we can plan for this.

Milan.

[ratewhore]

Still should be in the technical forum. Where would we be if everyone flouted the rules like you Milan?