Hi, Riyaj Shamsudeen schrieb: > Martin > As you correctly pointed out, PX slaved are blocked. > One obvious reason I can think is smaller large pool. What is your > large pool size set to ? Do you use ASMM - Automatic Shared memory > management stuff? If yes, do you see any growth or shrink for large pool > ? If large pool is not configured, it is possible that shared pool > fragmentation to cause PX slaves to wait for these PX events. v$sgastat > is a good place to start. Uh, the large pool approach sounds cool, I will check this ASAP. What I did so far: The SGA autotuning features are enabled (SGA_TARGET set), and shared_pool_size has been set to zero. I've added up the shared pool bytes from v$sgastat, and I ended up with about 120MB. Since RAM is no problem on this box, I changed SHARED_POOL_SIZE manually to 512MB recently (what, of course did not change anything for the running query described above. Yes, it's still running...). Why 512MB? I followed http://download.oracle.com/docs/cd/B19306_01/server.102/b14237/initparams197.htm#CHDGGAIB > What other events do you in statspack report or AWR report during that > time frame? Nothing that really matters, I just can't access the system at the moment, if important I'd have to post that on monday. Thanks so far, Martin -- Usn's IT Blog for Linux, Oracle, Asterisk http://www.usn-it.de -- //www.freelists.org/webpage/oracle-l