![]() There is no much concurrent access on those tables. After June, 30th (21.6): green line regularly bumps of +34 connections (July, 1st - July, 5th).As I have 2 CH, that means 16 connections per CH, which is the default value of postgresql_connection_pool_size June, 25th (21.4): green line is at 32 idle connections and concerns 1 table in 1 PostgreSQL database.The main difference I saw is the steady amount of idle connections I had in 21.4 vs the increase from the 21.6 version. I did not check the code, does PostgreSQL dictionary and PostgreSQL tables behave the same way from a connection point of view? If there is not much concurrent access - then as a quick solution can be to reduce value of postgresql_connection_pool_size and consider to use postgresql_connection_pool_wait_timeout to configure how much to wait on empty pool.Ĭurrently 5 tables (+2 views that uses some postgresql() functions). ![]() Also thinking about making one singleton pool for all postgresql tables. So I believe there should be the same issue in 21.4.Īnd the issue can remain, as I believe, only if there is concurrent access to tables, which fills all pool at some point and yes, this pool is not checked for having idle connections - I will add configurable keepalives_idle setting to postgres connection string, hope it will help. It is possible in case of concurrent access to tables, but onlyy in scope of one pool size.Īfter an update from CH 21.4.6.55 to CH 21.6.5.37Īctually a little strange, because 21.4.1 had also reported connections issue #23897 compared to 21.3 and then there were changes, which were not backported to 21.4. ![]() New connections to CH seem to trigger new connections to PostgreSQL
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |