Matthias Radestock

That's odd. I suspect the volume creates as a result of the [VOLUME declaration](https://github.com/docker-library/mysql/blob/ad625c64a06e16683e997e5a0147508d115f4989/5.7/Dockerfile#L71) ends up being read-only. Perhaps that's some strange OpenShift default permissions issue.

Are you sure that will solve the problem? The catalogue-db k8s manifest isn't mounting any dirs from the host into the container. Nevertheless, to experiment, I suggest creating your own...

> Is it possible to add IPv6 support? Sure. Shouldn't be hard. PRs welcome.

> But when servers work both with IPv4 and IPv6 connections establish only using single address family. Which one? I believe the general rule is that when there is a...

This is not a problem anybody should ever run into. If weave fails to remove entries for dead containers, we have a problem of much greater magnitude.

> But there's also no reason for a remove operation to not be idempotent. Sure. I didn't close this issue ;) Are all the other remove variants idempotent?

We do not know whether the container in question was gone completely or just not running. Are we saying that `weave dns-remove [ ...] [-h ]` should remove all DNS...

> dns records should be matched by container id, short-id, or name. We don't record container names in DNS.

> "should not" != "does not" I still think this is complete overkill. Don't forget that the user already has a way to clear any garbage: bounce the router. That...

We think the problem that gave rise to this was fixed in 1.4.6 (via #2023). I suggest this issue should make `weave dns-remove` idempotent in the same way as we...