I have a few instances where Docker containers work in combination (e.g. Nextcloud and Bookstack each work with mariadb; Telegraf, InfluxDB and Grafana all work together etc).

 

When there is a update for one or more of these Docker containers, is there a recommended way to update (e.g. stop everything, update everything, then restart in a specific order)? Or can I just update any individual container as and when I please without stopping any other "upstream" or "downstream" containers?

Depends. Well written apps will detect that a dependency isn't available and either retry or wait, then error out gracefully after a period of time. Poorly written apps may lose or corrupt data.

 

I've never had an issue with my specific set of containers, but ymmv.

19 hours ago, jonathanm said:

Depends. Well written apps will detect that a dependency isn't available and either retry or wait, then error out gracefully after a period of time. Poorly written apps may lose or corrupt data.

 

I've never had an issue with my specific set of containers, but ymmv.

When you say "well-written", do you mean on the part of the container creator or the underlying service? For example, I generally prefer to use the Linuxserver variant of a given container. Presumably, those would count as well-written?

 

Also, with regards to data loss, I imagine that depends on whether data is actively being written (e.g. my Nextcloud and Bookstack containers are usually NOT writing data whereas my telegraf container is constantly writing to InfluxDB)?

 

In any case, since I'm certainly not knowledgeable enough to know if an app is well written, I suppose I'll stick with my existing protocol of stopping all affected containers before updating. Small price to pay for peace of mind...