Re: enq: tx - contention on selects

Mark,

Thanks for the reply

> A select for update would require an ITL slot. Distributed
> transaction also take RBS entries.

These are just plan selects (ie NO for update clause), but they are part of distributed transactions. At any time, there may be about 40 of these distributed transactions occurring at once (which isn't really a massive number). We are using undo tablespace (as opposed to Rollback segments) - are there a limited number of slots available for undo? Could this TX - Contention wait be caused by a transaction attempting to get a slot of undo?

>
> A share mode of 4 usually results from no free ITL slot available, a
> unique key insert, or a wait to update a bitmap index.

I know we have no bitmap indexes in the application and as the waits are against a select, I am still confused as to why I am seeing the TX Contention at all, unless its something to do with undo.
>
> Also verify that the select is not going against data affected by
> updates to the partent in a FK relationship where no index exists to
> support the FK.

I don't think this is the case, but I will check for sure.

Thanks,

Stephen. Received on Tue Sep 16 2008 - 09:10:23 CDT