Unanswered: memory config for ms failover cluster sql
here i got a little doubt for my cluster sql memory config...
the server's ram is 2 gb, and by default, sql server's max memory is 2 gb. my question is, do i need to reduce the sql's max memory in order to make sure the OS will run smoothly? if yes, how much reduced? 128 mb?
the reason for my concern is that my cluster sql server is heavily loaded, i am afraid it may approach the max memory for some moment, then the windows server crashed ... especially for a cluster environment (active/active), if failover does happened, at one shot the memory will be all used for sql server ... is it possible?
Sql Server 2k generally performs reasonably well in terms of sharing resources with the OS. If the cluster is to support non - Sql Server applications in a failure situation, one (moderatly conservative) approach may be to add up all of the minimum RAM requirements for the remaining applications and set the maximum setting such that at least that amount of RAM would be available to the other applications in the event of a node failure. (However, it would be best to perform node failure load tests, so that one may unambigously know what kind of performance to expect in node failure situations under nominal and peak production load conditions.)
I believe that if you leave everything the ram settings defaulted, that the OS takes its share of the ram first, then SQL takes what is left. You should be able to fail over without a hitch, in ram terms anyway. I would certainly test it out though.