This feels like heavily burying the lede. It looks to me like v1.49 will move docker nodes over to the storagenode-updater and include them in phased rollouts.
This just doesn’t stand out in the diff because the entrypoint file has moved in the repo. But this will be a very impactful change that honestly has me slightly worried whether everything will keep working as intended.
For one this means that the same docker image can suddenly result in a different node version. Which isn’t usually how you would expect docker images to work.
I have quite a few questions regarding this too. How will this work in conjunction with watchtower? Does this system prevent inadvertent downgrades which could cause problems after database migration to a new version? Plus there are people with manual update scripts. Could that cause issues and can they still use those? (I’m guessing no)
I know I will want to keep a closer eye on things with this update. To aid with that, would it be possible to tell us when the new docker images are going to be available before it happens?
We still have issues with the storage node updater inside docker. We reverted it on the release branch once more. Maybe we have a working solution in the following version.
If you try to downgrade a storage node it will refuse to start. That is the protection we have in place. For the storage node rollout we will never signal an old version. If we encounter an issue we will fix it and release a point release. That point release might revert what the previous version has broken but it would still be a database upgrade and never a downgrade.
I assume they would upgrade the docker image but even the new docker image would still follow the rollout process and might install the old storage node version. These people would only update the updater. Old and new updater would behave the same way.
Yes sure. Best would be to test this in our testnetwork
