swarm mode for VIM

[homework36]

For production, we would like to have containers self check and restart if needed. Therefore, we will need a YML file for the Docker swarm mode.

[dchiller]

Checks and restarts are available with docker compose in non-swarm mode (eg healthcheck)... I think swarm is for multiple nodes?

[homework36]

Checks and restarts are available with docker compose in non-swarm mode (eg healthcheck)... I think swarm is for multiple nodes?

Thank you! I wasn't aware of this. Do we have checks and restarts implemented already? Sometimes our containers will fail (see logs below), which results in 502 nginx error, and we have to manually restart the container.

In addition to the setting in docker compose, I think the current prod instance might be too small with 1 vCPU and 1.5 GiB RAM. Should we resize for a larger one?

[dchiller]

Do we have checks and restarts implemented already?

Looks like we do for some of the services and not for others (see here. The postgres container has a healthcheck and will restart automatically based on the unless-stopped rule. Looks like nginx will also restart. The django app container doesn't restart currently.

[homework36]

Would it make sense to consider adding healthchecks for the remaining containers as well?

Do we have checks and restarts implemented already?

Looks like we do for some of the services and not for others (see here. The postgres container has a healthcheck and will restart automatically based on the unless-stopped rule. Looks like nginx will also restart. The django app container doesn't restart currently.

[dchiller]

Would it make sense to consider adding healthchecks for the remaining containers as well?

Adding a restart instruction to the app container seems like it would be useful based on what you are describing, although it's probably also worth figuring out what's going on to cause those timeouts.

Perhaps there are relevant health checks, too...what they are depends on what you want to monitor.

[fujinaga]

In addition to the setting in docker compose, I think the current prod instance might be too small with 1 vCPU and 1.5 GiB RAM. Should we resize for a larger one?

If possible, yes. I'd like UMIL to be responsive as possible.