On 2/4/19 7:33 AM, PegoraroF10 wrote:
About replication ... Logical Replication with CREATE
PUBLICATION/SUBSCRIPTION.
Yes, some DDL commands were ran on that server but none of them were related
with that select.
Let me explain better. We have a single server with a single database on it.
Each customer has its own schema and connects to it to work exclusively on
that schema. So, sometimes we add some customer or change something on an
old one. But this DDL change we could ran is not related with that schema we
are talking. Schemas can have different structures but that schema which
puts my server on recovery mode was not changed.
Maybe something related happened some days ago. When we start a new customer
we add a schema, put all their tables on it and it´s ok. Our server has
today 90 schemas and each schema has 100 tables, resulting in 9000 tables.
Some days ago we added 5 new customers on same day, so we added 500 tables.
Then, when we did that some selects on system tables were very very slow and
that was only solved when we did a REINDEX DATABASE. Even REINDEX SYSTEM did
not solve. Is this problem related with recovery mode of my server ?
Unsure at the moment as there is not enough information to come to any
conclusions.
Questions:
1) Do you still have the logs from when you added the 5 new schema and
do they show any warnings, errors, etc at or after that time?
2) Are you seeing any warnings, errors, etc in the logs currently?
3) Is the replication for all tables?
4) The exact same query works without a problem on the replicated
server, correct?
5) In the Postgres server that is being replicated to are there any log
entries that might be of concern?
6) Is it possible to get a stack trace of the crash?:
https://wiki.postgresql.org/wiki/Getting_a_stack_trace_of_a_running_PostgreSQL_backend_on_Linux/BSD
--
Sent from: http://www.postgresql-archive.org/PostgreSQL-general-f1843780.html
--
Adrian Klaver
adrian.klaver@xxxxxxxxxxx
