CouchDB not starting with default parameters

[guanana]

---

BUG REPORT

Description

swarmpit does not start. It seems to be a problem with couchdb container

Steps to reproduce the issue:

1. Try to install it
2. Couch DB gets stuck in 137 exit error
3. Increase mem limits on CouchDB and then the app cannot connect

What happens:

The app never starts since couch db is NOT available

What should happen:

CouchDB should start properly so the app can connect and start using the DB

Additional information (e.g. docker version, cluster setup,...):

I solved the problem by adding manually the following on the deployment file:

  db_couch:
    image: couchdb:2.3.1 <-- I as well updated to latest bug fix version from 2.3.0 to 2.3.1 just in case that made a difference but not
    volumes:
      - db-data:/opt/couchdb/data
    networks:
      - net

Section separated for clarity on what was added:

    ulimits:
      nofile:
        soft: 1000
        hard: 2000

    deploy:
      resources:
        limits:
          cpus: '0.90' <-- Increased cpu to see if that made a difference... it didn't
          memory: 2048M <-- Increased mem

Before that the DB was just hanging and the app unable to connect saying Connection refused.

Waiting for more than 10% CPU topping up and nothing happened.

Change the ulimits for the couchdb container and I get a healthy connection almost immediately.

[heatzync]

I also experienced this and @guanana's solution above is the simplest solution.

I'm adding my symptoms here for anybody else that might encounter the same...

Symptoms

• The swarmpit_db service kept restarting after a couple of seconds (closer to every minute)
• The swarmpit_app service was complaining that it couldn't reach http://db:5984
• The swarmpit_db CPU usage was 100%
• There were no logs for swarmpit_db except for the warning about "Admin Party" being printed each time a new swarmpit_db container was started

Bad solution

I suspected the CouchDB container wasn't getting enough resources, so I commented out the deploy.resources section in the swarmpit/docker-compose.yml file to see how much it needs:

#    deploy:
#      resources:
#        limits:
#          cpus: '0.30'
#          memory: 256M
#        reservations:
#          cpus: '0.15'
#          memory: 128M

The memory usage climbed to ±1.5GB and the CPU was still 100%, but the swarmpit_db service wasn't restarting any more, so I left it running for a while. It eventually started successfully after ±25 minutes and the CPU went down to less than 10%. I thought that was the solution, but I wasn't satisfied.

I also tried bumping the CouchDB version to 2.3.1, but the same symptoms persisted - it only starts after using ±1.5GB memory and 100% CPU for many minutes.

I kept searching the webs and found this reported issue with @guanana's solution! 🕺