Re: PostgreSQL logical replication depends on WAL segments?

Hi guys, thank you very much for all information. I learned my lesson
regarding cronjob cleaning old WAL logs...

There is one other interesting problem I have found today and I would like
to ask you about you opinion.
On logical master I found this morning big flood of these messages in
postgresql log:

2019-01-23 08:17:42.338 UTC [1310] WARNING: oldest xmin is far in the past
2019-01-23 08:17:42.338 UTC [1310] HINT: Close open transactions soon to
avoid wraparound problems.
You might also need to commit or roll back old prepared
transactions, or drop stale replication slots.

Since this is still testing environment I dropped subscription and
publication and logical replication slot and messages stopped.
But do you know what was wrong? Why these warnings? Did logical replication
slot blocked autovacuum or something else happened?
Can I prevent it?
As mentioned before we have quite huge amount of data, around 500 millions
of records every day, most of them inserted in separate transactions or in
only small blocks so I guess xmin value can grow very quickly...

Thanks

jm

On Tue, 22 Jan 2019 at 20:03, Jeremy Finzel <finzelj(at)gmail(dot)com> wrote:

> Note replication slots only prevent old *catalog* rows from being
>> removed, not old row versions in user created tables.
>>
>
> Thank you for that clarification. I can see this is noted clearly in the
> CAUTION statement here:
> https://www.postgresql.org/docs/current/logicaldecoding-explanation.html#LOGICALDECODING-REPLICATION-SLOTS
>
> Thanks,
> Jeremy
>