Re: [Glue] Deadlock bug

Josh Berkus  wrote:
>> the behavior was the same up to the second UPDATE on Process 2, at
>> which point there was no deadlock.  Process 2 was able to commit,
>> at which point Process 1 failed with:
>>  
>> ERROR:  could not serialize access due to concurrent update
>
> Does this happen immediately, not waiting 2 seconds for deadlock
> checking?
The deadlock checking delay never comes into play.  Process 2 would
never be blocked, and Process 1 would fail on the COMMIT of Process
2.
Without a detailed scenario I can't comment on exact behavior, but in
a serializable-only environment, with SSI enforcement of RI, you can
count on only having blocking on write/write conflicts, so it would
only be a cycle of those which could ever cause a deadlock.  Anything
where deadlocks currently occur because of SELECT FOR SHARE or SELECT
FOR UPDATE would not have the same deadlock issues.
-Kevin