ted with vacuum skip locked pages patch + freeze +
shared catalog combination, but cannot prove it yet.
Looking for possible course of action.
Probably simplest fix - drop and recreate these 6 affected users, but so
far I willing spent some time research into this issue.
--
Maxim Boguk
S
:
select relfrozenxid from pg_class where relname='pg_authid';
relfrozenxid
--
2863429136
So it's interesting where value of " from before relfrozenxid 248712603"
come from.
--
Maxim Boguk
Senior Postgresql DBA
http://dataegret.com/ <http://www.postgresql
On Tue, May 22, 2018 at 9:47 PM, Andres Freund wrote:
> Hi,
>
> On 2018-05-22 21:30:43 +0300, Maxim Boguk wrote:
> > For sample:
> >
> > postgres=# vacuum pg_catalog.pg_authid;
> > ERROR: found xmin 2894889518 from before relfrozenxid 248712603
> >
>
On Tue, May 22, 2018 at 10:30 PM, Andres Freund wrote:
> Hi,
>
> On 2018-05-22 22:18:15 +0300, Maxim Boguk wrote:
> > On Tue, May 22, 2018 at 9:47 PM, Andres Freund
> wrote:
> > > > select relfrozenxid from pg_class where relname=
gs,
>
> Andres Freund
>
Ok this issue reproducible on promoted replica.
So now I have my playground.
Will provide gdb info in few hours.
--
Maxim Boguk
Senior Postgresql DBA
http://dataegret.com/ <http://www.postgresql-consulting.com/>
Phone RU: +7 985 433
Phone UA: +380
On Thu, May 24, 2018 at 12:38 PM, Maxim Boguk wrote:
>
>
>>
>> > About gdb bt - it's tricky because it is mission critical master db of
>> > huge project.
>> > I'll will try promote backup replica and check is issue persist there
>> and
>
sages
for the same prepared statement."
Hard to say what the reason for slow planning, but one (there could be
others) likely reason is JIT work. Do you have JIT enabled?
--
Maxim Boguk
Senior Postgresql DBA
https://dataegret.com/
Phone UA: +380 99 143
Phone AU: +61 45 218 5678
ng stage slowdown.
Can you run explain (analyze, costs, buffers, timing) of your query?
And check how long the planning stage took and check if JIT was used or not
(and how much time spent during JIT if it had been used).
--
Maxim Boguk
Senior Postgresql DBA
https://dataegret.com/
Phone UA: +3
age of almost 200x of work_mem very surprised me).
PS: after crash recovery of a database with a few millions of files in
pgsql_tmp takes ages (however it is fixed in newer versions so it's not a
critical issue).
--
Maxim Boguk
Senior Postgresql DBA
https://dataegret.com/
Phone UA: +380 99 143
Phone AU: +61 45 218 5678
-
> Bitmap Heap Scan on common_student (cost=4.91..243.76 rows=64 width=385)
> (actual time=0.128..0.267 rows=50 loops=1)
>Recheck Cond: (school_id = 1)
>Heap Blocks: exact=16
>-> Bitmap Index Scan on "unique common_student" (cost=0.00..4.90
> rows=64 width=0) (actual time=0.110..0.110 rows=50 loops=1)
> Index C
10 matches
Mail list logo
