[rescue] Trimming down Solaris 10
velociraptor at gmail.com
Fri Dec 5 14:13:47 CST 2008
On Fri, Dec 5, 2008 at 9:31 AM, John Lengeling
<John.Lengeling at radisys.com> wrote:
>> (And thanks to Francios for the further suggestions on Sol 10
>> packages. I'd like to convince them to use ZFS; we'll see how far
>> that flies given they just bought "new" V490's since they don't think
>> T5120's can "handle the database load".)
> What 32 cores in a T5120 aren't enough? How many are in a V490?
Don't get me started.
There are two M5K's sitting collecting dust because they didn't think
the 6CPU version would be able to handle what the 3 V490's, and 4
T5120s are carrying (1 part of the db cluster with the V490s, the
others in an N+1 cluster running MQ on VCS).
They think the M5K's will "only" be able to handle the data warehouse
instance once it gets spun up on and "too big" for the V490, so they
haven't been built and deployed yet.
Seriously, clues are needed by my boss(es)^2. Due to the deadlines,
I've not really had the cycles to disabuse them of these notions. We
(dba and I) do plan to have a "bake off" between the T5120 and the
V490 in he db cluster by moving the event logging db the T5120 will
normally run onto a V490, and putting the customer db onto the T5120
and then stress testing. Our existing 3-node V490 (customer dbs) + 1
T2K (also doing event logging) clusters show the V490's getting pegged
probably 20-30% of the time.
Most of our load issues boil down to being (essentially) DDoS'd by our
own customers. You can take some guesses as to why, as I really
shouldn't get into the details on a public list.
More information about the rescue