Hi,
the perf profiles look really weird - clearly, there's a lot of lock
contention, because the top item is this
13.49%13.40% postmaster postgres [.] LWLockAcquire
|
---LWLockAcquire
That's a sign of lock contention - not sure which one,
Hello, any news ?
Thank you,
Alessandro.
On Fri, Mar 23, 2018 at 8:22 PM, Alessandro Aste
wrote:
> PS , in the meanwhile I discovered a 2nd workaround(beside disabling
> parallel processing) . I added offset 0 to the subquery , and, according
> to the
PS , in the meanwhile I discovered a 2nd workaround(beside disabling
parallel processing) . I added offset 0 to the subquery , and, according
to the documentation, “OFFSET 0 is the same as omitting the OFFSET clause” -
https://www.postgresql.org/docs/current/static/queries-limit.html
cmd3dev=#
Thanks Tomas. We're currently building postgres from source. In order to
enable symbols, you want me to re-configure postres with --enable-debug
then run perf?
Regards,
On Thu, Mar 22, 2018 at 5:00 PM, Tomas Vondra
wrote:
>
>
> On 03/22/2018 11:33 AM, Alessandro
On 03/22/2018 11:29 PM, Alessandro Aste wrote:
> Thanks Tomas. We're currently building postgres from source. In order to
> enable symbols, you want me to re-configure postres with --enable-debug
> then run perf?
>
Yes.
regards
--
Tomas Vondra http://www.2ndQuadrant.com
On 03/22/2018 11:33 AM, Alessandro Aste wrote:
> Tomas, thank you. This machine is abare metal server running only a
> staging postgresql 10.3 instance. Nobody is using it beside me.
>
> I'm attaching 4 files.
>
> every_30_seconds_top_stats_during_query.txt - this is a caputure of
>
Tomas, thank you. This machine is abare metal server running only a
staging postgresql 10.3 instance. Nobody is using it beside me.
I'm attaching 4 files.
every_30_seconds_top_stats_during_query.txt- this is a caputure of the
top command every 30 seconds(more or less) for 10+ minutes while
On 03/21/2018 08:44 PM, Alessandro Aste wrote:
> Thanks for your reply Tomas. The query just got stuck for forever. I
> observed no CPU spikes, it is currently running and I see 89 of the CPU
> idle. I'm using a relative powerfull machine (120 GB ram, 56 CPU).
>
That doesn't really answer
Thanks for your reply Tomas. The query just got stuck for forever. I
observed no CPU spikes, it is currently running and I see 89 of the CPU
idle. I'm using a relative powerfull machine (120 GB ram, 56 CPU).
Plain analyze as requested. :
QUERY PLAN
On 03/21/2018 05:09 PM, Alessandro Aste wrote:
> Hi there, we are using postgresql 10.3 and we're facing an issue with a
> query. The query (full query below) completes only when:
>
> 1 - LIMIT 10 is removed
> or
> 2 - show max_parallel_workers_per_gather is set to 0, so parallel
>
Hi there, we are using postgresql 10.3 and we're facing an issue with a
query. The query (full query below) completes only when:
1 - LIMIT 10 is removed
or
2 - show max_parallel_workers_per_gather is set to 0, so parallel
processing is disabled.
With max_parallel_workers_per_gather set to
11 matches
Mail list logo