On Thu, Nov 25, 2021 at 10:53 AM James Sewell <james.sewell@jirotech.com> wrote:
>
>
>> > Seems like some of the processes are taking a long time or stuck while
>> > reading/writing SLRU pages, and due to that while creating a new
>> > connection the backend process is not able to check the transaction
>> > status (from pg_xlog) of the pg_class tuple and gets stuck/taking a
>> > long time in a startup.
>>
>> My next question would be whether this particular process shown is
>> stack is stuck forever or finally, it is able to make a connection? I
>> want to understand is this just due to slow I/O or something else?
>
>
> They don't ever seem to clear, (more and more show up) and IO doesn't seem to be deadlocked at that time.
How did you verify that, maybe some process started IO and stuck
there? Can we check pg_stat_activity that is there some process that
shows in the wait event as SLRURead/SLRUWrite and not coming out of
that state?
--
Regards,
Dilip Kumar
EnterpriseDB: http://www.enterprisedb.com