sga_max_size doesn't have to be the same for all instances so you can change it on a per-instance basis, but it requires an instance restart to take effect. You could stop one instance at a time using "srvctl stop instance -d orcl -i orcl1" and then restart it before restarting the other instance. For well-behaved applications, that shouldn't cause any outage (note: many applications are not well-behaved). Dan On Thu, Jan 7, 2010 at 12:44 PM, K R <kp0773@xxxxxxxxx> wrote: > so suppose i want to change the sga_max_size for a 2 node RAC database. so > will it work if i only restart one instance . normally for all the instance > i use a srvctl stop database command. > > > On Thu, Jan 7, 2010 at 8:41 AM, Dan Norris <dannorris@xxxxxxxxxxxxx>wrote: > >> Hi Scott, >> >> Technically speaking, a database can't start. A database is just the >> datafiles, control files and redo logs. An instance is the shared memory and >> background processes, so that's the part that gets started. (BTW, I've never >> been pleased with the "srvctl start database ..." syntax in this regard.) >> >> The gv$instance view is the right place to look. However, if there are >> multiple start/stops for the instances (and users get "ping-ponged" between >> instances), you couldn't tell from there alone--you'd have to look at alert >> log information. >> >> Hopefully, I've understood your question properly. >> >> Dan >> >> >> On Thu, Jan 7, 2010 at 9:48 AM, Scott Sibert <ssibert@xxxxxxxxx> wrote: >> >>> I've looked around at a few things but we're pretty new running RAC and >>> I'm not sure where to find this information. >>> >>> I'm trying to find the start time for the database, not the individual >>> instances. For example, INST1 and INST2 start Monday. Tuesday INST1 >>> restarts, thus causing gv$instance for INST1 to show Tuesday as start time. >>> Then, Wednesday, INST2 restarts, showing INST2 start time as Wednesday. >>> >>> So far, we have INST1 start time as Tuesday, INST2 start time as >>> Wednesday, but the database itself has been up continuously since Monday. >>> Where do I find that Monday start time since Monday is the true start time >>> of the database? >>> >>> Thanks. >>> Scott >>> >> >> >