Re: Deadlocks with block - 0, slot - 0 during INSERT

On Jul 10, 12:39 am, ca111..._at_gmail.com wrote:
> Hi
>
> We are getting strange deadlocks in 9.2.0.7 database:
> 1. They occur during INSERT into range-partitioned table where all
> indexes are local.
> 2. Both sessions were trying to execute the same SQL statement.
> 3. ObjectID point to table partition and index (primary key) partition
> corresponding to this table partition. This is strange as I would
> expect locking to occur in the same order in both sessions.
> 4. "Rows Waited On" section of trace file shows block - 0 and slot -
> 0:
>
> - - - - - - - - - - - - - - - - - - - - - -
> Rows waited on:
> Session 65: obj - rowid = 00009B64 - AAAJtkABsAAAAAAAAA
>   (dictionary objn - 39780, file - 108, block - 0, slot - 0)
> Session 90: obj - rowid = 00009B65 - AAAJtlAASAAAAAAAAA
>   (dictionary objn - 39781, file - 18, block - 0, slot - 0)
> - - - - - - - - - - - - - - - - - - - - - -
>
> Why are block/slot = 0?
>
> In other databases we get something like that:
> - - - - - - - - - - - - - - - - - - - - -
> Rows waited on:
> Session 513: obj - rowid = 000068C1 - AAAGjBAALAAAABHAA1
>   (dictionary objn - 26817, file - 11, block - 71, slot - 53)
> Session 468: obj - rowid = 000068C1 - AAAGjBAALAAAAA9ABM
>   (dictionary objn - 26817, file - 11, block - 61, slot - 76)
> - - - - - - - - - - - - - - - - - - - - -

See http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:1528515465282 and search for:
silly
left-overs

That and other deadlock threads on asktom and oracle-l can be quite informative.

jg

--
@home.com is bogus.
"People often say treat others as you would like to be treated
yourself. Does not work like that in technical forums. Treat an expert
like a newbie and you will offend. Treat a newbie as an expert and you
only make an already frustrated person even more frustrated." - Billy
Verreynne