I’ve recently updated from 1.6 to 220.127.116.117. I’ve just done my first Tentacle upgrade after updating the server to 18.104.22.1680 and noticed some strange behaviour.
We use Octopus in an unusual way in that our environments map to clients, so each client has their own instance of the application. In 1.6 I was able to call each machine on the single physical server the same name, and they have the same IP address. In moving to 2.0 I needed to individually name the machines, but they’re still the same physical server.
What seems to happen in the Tentacle upgrade push is that the Octopus server doesn’t notice that a large number of the machines are actually the same tentacle, with the same thumbprint and same IP address. It therefore schedules 20+ tentacle upgrades, queing the last 19 of them whilst the first one stops, upgrades and restarts, then the next one, repeat 18 more times. This obviously ties up the server far longer than ideal. It would be quicker for me to do the upgrades manually as I only have 5 physical servers, but 50+ machines across a few projects.
I’m currently 28 minutes into the upgrade and it looks like about 1/6th of the ‘upgrades’ are done.