Hi,
On 2018-04-04 11:54:14 +0300, Konstantin Knizhnik wrote:
> Several times we and our customers are suffered from the problem that
> Postgres got stuck in deadlock detection.
> One scenario is YCSB workload with Zipf's distribution when many clients are
> trying to update the same record and compete for it's lock.
> Another scenario is large number of clients performing inserts in the same
> table. In this case the source of the problem is relation extension lock.
> In both cases number of connection is large enough: several hundreds.
Have you ever observed that in the field? This sounds more artificial
than real to me.
> So, I see three possible ways to fix this problem:
> 1. Yury Sololov's patch with two phase deadlock check
> 2. Avoid concurrent deadlock detection
> 3. Avoid concurrent deadlock detection + let CheckDeadLock detect all
> deadlocks, not only those in which current transaction is involved.
4) Don't do anything about deadlock detection, because this is just a
symptom five steps removed from the cause.
We've to pay attention to fixing actual problems, rather than purely
benchmark ones. Complexity has a significant price.
Greetings,
Andres Freund