copy-n-paste from submitted suggestion weeks ago
Suggest for future upgrades, and more so for major ones is a 2 pronged approached.
- 2 step upgrade,
- 2 step deploy,
2 step upgrade
First is an upgrade to an intermediate version that operates seamlessly with the current version and will work seamlessly with the final new version.,
This allows time for people to upgrade to the intermediate and when enough people have upgraded then the final version can be rolled out.,
This also means that major changes that do not work so nicely with the old version will not require a sudden upgrade to occur and disrupt and cause large churn events.,
We still have people running hundreds of thousands of nodes who have a script that brings down all their nodes, deletes their directories, get new version and then start them afresh.,
2 step deploy
This is where a “trusted” group of people on an opt in basis is given the upgrade to upgrade to (yes 2 step upgrade for major upgrades),
They test and use the new nodes on one of their machines to judge its stability on the mainnet.,
A week later if all reports are good, or manageable then deploy to everyone.,
For major upgrades then this is a double process, although minor ones where the upgrade still works seamlessly then the 2 step upgrade usually will be unnecessary.