Re: Locking Question on OPS??

ccdicky wrote in message <80vfkd$76m1_at_imsp212.netvigator.com>...

>I queried several system views, including v$lock, v$system_wait .. on both
>nodes but cannot find any clue.

Those V$ tables are local to that instance - they do not give any indication about conetntion between nodes. You need to look at the OPS specific systems tables. There's a couple of V$ views that provide cross-instance data - i.e. data about the entire database and not just the local instance.

With Oracle 8 you also get the new global views that can be used - they provide even more information about the entire database.

Not sure what type of DLM (Distributed Lock Manager) you're using. With OPS 7.x that was a 3rd party software written by the UNIX vendor. The DLM handles cross instance locks. IOW, process on Node A wants to update record xyz, but there's a read only lock on xyz put in place by a process on Node B. Node A will then ping the DLM, request the lock, the DLM will ping Node B, requesting it to degrade the lock so that the contention is resolved, upon which Node B will ping the DLM with the result and the DLM pinging Node A again.

So if lock contention is a problem, and you do not think it is not due to the clients' SQL statements, it may just be that you ran into either a bug or a misconfiguration in the DLM.

Lastly, are you by any change 1using PQ? If you are, the the process on Node A may have PQ slaves running on Node B. This should not be a problem, but then hey, I've seen weirder things in Oracle. :-)

regards,
Billy Received on Thu Nov 18 1999 - 00:47:36 CST