Hi Experts,
Your input is most welcome!
We are using Postgres 13 (and plan to upgrade to 15 soon). We have logical replication with about 40 servers subscribing to one publisher. 40 Walsender processes are running on the publisher server. When we insert a row into a table holding binary data the walsender RAM usage increases by 500MB although the row binary data is only 80MB. We see this increase in all walsender processes. At some point we got OOM and the process was killed.
- Why does the walsender increases by 500MB when the data change was only 80MB
- Is some of the 500MB increase due to shared memory or each walsender has its own 500MB increase. I assume that if it was only in shared memory we would not have gotten OOM…
- Why when logical_decoding_work_mem = 64MB the RAM is 15 times that size? Shouldn't any additional space be used from disk and not RAM?
- Will adding streaming = on to publication "PUBLICATION pub WITH (streaming = on)" can alleviate the issue?
- Are there configuration options that can resolve the RAM issue. It can be also in version 15 since we plan to upgrade soon.
Thanks!
IMPORTANT - This email and any attachments is intended for the above named addressee(s), and may contain information which is confidential or privileged. If you are not the intended recipient, please inform the sender immediately and delete this email: you should not copy or use this e-mail for any purpose nor disclose its contents to any person.