Hi there, I’m seeing an issue on my docker install. It’s on Linode on Ubuntu 14.04. Postgres is using 100% of the CPU. The forum itself is working just fine, version v1.6.0.beta7 +156. Doesn’t seem any slower or anything. Here’s a screen capture from “htop”: [תמונה] Any pointers on how I can diagnose this and remedy it? I’ve tried rebuilding with ./launcher rebuild and even cleaning up with ./launcher clean. I’ve tried rebooting the server as well. I have plenty of disk space left (10GB free), and Discourse is the only thing running on this server. Thanks in advance for any pointers.

Replying to my own post, because I realized my mistake :slight_smile: VACUUM VERBOSE ANALYZE only works if I select the discourse database, of course :slight_smile:

Are there any jobs running? Try opening the following URL for your site: /sidekiq

Doesn’t look like anything else is running. [תמונה] [תמונה] [תמונה] [תמונה]

Ok, there is one job in the “scheduler” tab that is running. Jobs::EnsureDbConsistency 41 minutes ago RUNNING -1ms _scheduler_localhost-app:29119:1:29c83b7cc82551f6a2b801e82a1acb38 [תמונה] Should I just wait that out? Seems like it’s checking the database?

To see exactly what queries are being executed you can run something like the following from the console: cd /var/discourse/ ./launcher enter app sudo -u postgres psql -c "select * from pg_stat_activity WHERE state='active'"; Try running this a couple of times with a 3/4 minute wait in-between, capturing the output each time.

Ok, looks like it’s running 3 copies of this query (with 3 different pids) DELETE FROM user_actions USING user_actions ua2 WHERE ( user_actions.action_type = ua2.action_type AND user_actions.user_id = ua2.user_id AND user_actions.acting_user_id = ua2.acting_user_id AND user_actions.target_post_id = ua2.target_post_id AND user_actions.target_post_id > 0 AND user_actions.id > ua2.id ) Full copy/paste here: datid | datname | pid | usesysid | usename | application_name - Pastebin.com I’ll let it run for a while and see what happens. Maybe I’m just being impatient :slight_smile:

Yes, you’ll need to let those queries run, because otherwise the scheduler will just keep spawning them and you’ll be playing whack-a-mole with them forever.

Do they start every x minutes or something like that? I currently have 5 of them running … Up from 3 a few hours ago.

Well, this morning we’re up to a load average of 12 :slight_smile: [תמונה] But it does appear to have moved on from the “DELETE” queries, and is running others. I assume this is some sort of database check. The forum is still responsive with this load, so I’ll let it proceed for a while. Seems a bit crazy though.

Are you running on our latest docker image? How many posts does your forum have?

Yes, I updated the docker image yesterday, which is when this started happening. I did git pull ./launcher cleanup ./launcher rebuild app ./launcher cleanup As for posts and topics, it has 17001 topics, with 178297 posts. They were imported from a phpBB install last year. Anyway, it eventually got up to 16 instances of that query running, so I just killed them off. They haven’t returned in a few hours. I’ll check if it starts using CPU when that scheduled task starts again.

Postgres using 100% of CPU

Product Support Self-hosting

MiG 5 ביוני,‏ 2016,‏ 12:28am 3

Doesn’t look like anything else is running.

נושא		תגובות	צפיות
Postgres has 100% CPU for large databases, Discourse 2.7.7 Self-hosting server-resources	3	862	30 באוגוסט,‏ 2021
Some queries that don't finish Product Support	16	1142	5 באפריל,‏ 2019
Primary Postgres database process (postmaster) eating all CPU Self-hosting server-resources	13	4542	28 במאי,‏ 2020
Slow Sidekiq + Postmaster using 95%+ CPU (32 cores) after Postgresql Version Upgrade Self-hosting server-resources	22	3645	29 במאי,‏ 2020
Unusually high CPU usage Self-hosting	31	1513	18 בפברואר,‏ 2026

Postgres using 100% of CPU

נושאים קשורים