I remember long time ago there was a strict rule, where all nodes should be updated within 72 hours after update released. If not, reputation went down or smth like that.
But now looks like there is no this rule anymore?
while the software was in alpha and beta yes 72 hours was the case but once it went into production updates to the software became a controlled rollout to prevent bad things that may not show their face until it went into real world environment hope I got the explanation correct.
well you still get warning emails… tho only after you are like 1 version behind…
i still think the update schedule is a bit strict, like now you can run 1.3.3 as the oldest version… ofc since updates are coming in frequently, it makes sense to try and keep nodes on the same version almost…
but it is nice when one only has to update like once every quarter… but yeah … future thing i suppose… atleast if i keep my manual updates… might go to watchtower if i end up trusting it
I believe that guideline is still similar, but it’s just 3 days after the rollout has finished now.
But there is some nuance, it’s not “If your node hasn’t updated three days after rollout is done, your reputation will be harmed”, but rather “if your node updates before three days after rollout is done its reputation won’t be harmed”. Beyond that 3 day limit, you may still be fine, but no promises.
I most cases you have much longer to update your node. But this rule gives storjlabs the option to update nodes to a later version fast should some change require it.
From what I can tell that reputation impact currently just means outdated nodes won’t be selected for uploads anymore. And possibly don’t get downloads either, though not entirely sure about that one. Though this may obviously change in the future. I can imagine the node would just be considered as offline at some point, which would hurt your uptime reputation.
3 days seems kinda of low to shutdown, updated and restart like 10000 nodes +
if we say 6k nodes which is very low estimate, which we divide by 24, so thats how many they would have to shutdown, updated and restart in 1 hour, so 250 and divided by 60 minutes in an hour so 4.16 nodes that would have to be rebooted
i guess if we assume three times the nodes and thus three times the time… then we end up with the same… 4.16 nodes a second for 3 days straight to update them all… i duno how many the system can do or would loose … but it does atleast in theory look is a pretty difficult task.
and then on top of that comes the whole… OPS we updated our entire network to a bad release and now every node crashed and we cannot restart any of them…