Gc cr multi block request.

Question

Hi ,
I am sufferiung from a 40 hour long running job which normally completes in few hrs, that jus creates a table (actually create table as select,joining many tables using parallel query withe 16 sessions).
There is only 2 session that are active at this time, one of them is the parent waiting on PX:DEQ EXECUTE REPLY.
The other sesion is running from 24 hrs, it is actually not waiitng upon any event for a signifcant time.Most of the time it faces GC CR MULTI BLOCK REQUESTS wait event. Is this due to interconnect latency?.I guessed it because the protocol used by the interconnect is UDP which doesnot wait for an acknowledgement. Can you please suggest me how to check for the latency on the interconnect?? Is my doubt on the interconnect is reasonable???????????
I checked the trace files,awr reports,ash reports and every thing but nothing seems to be working fine. All I see is that single sesion using up 98% of the CPU and the TOP wait event in the awr,ash reports being the CPU Time.Pleas help me.Thanks.

Answer 1

You are difficult to help as you don't post
- a four digit database version
- information on your OS
- information, including configuration info, of your interconnect
RAC is a chatty protocol, and sends large packages, of even 64k and higher. This means if you don't use an 1 Gb network card, and don't have 'jumbo frames' enabled, performance will suffer.
As a result, at least in 9i, you will see this very wait event show up all the time, at least on AIX.
Sybrand Bakker
Senior Oracle DBA