Comment 43 for bug 1752411

Revision history for this message
Trent Lloyd (lathiat) wrote :

I agree with the sentiment that 5 seconds feels too long, however as a workaround I decided I would just copy the existing timeout. I certainly would not want to make it longer since this is in the critical boot path.

I would generally agree that in general a DNS request should fail faster however there are some cases where it won't, e.g. spanning tree bring up on ports can take 2 seconds.

My hope is to correctly fix host after getting this in, since the impact is very high for affected users.

This check may actually be able to go away, I believe both systemd-resolved and libnss-mdns (latest version that I think is not in bionic) implement the .local label checking to do this at runtime instead of this old hack. So for cosmic+ we can probably get rid of this logic, which always sucked anyway. As we only needed to really disable nss-mdns and not avahi entirely (since apps should normally resolve the IPs using avahi's API anyway, the impact to actual avahi usage is low).

Since the impact is high but only on a smaller subset of users, I think we should go with matching the current timeout for now and worry about further improvements later.

I've verified the cosmic upload is working as expected on a non-affected system.