I had this idea in the shower, and I don’t even know if it would benefit the storj satellites or clients.
As a node operator, sometimes I’m taking a node down intentionally for maintenance. Sometimes it’s quick and sometimes it’s slow (migrating disks)
Would it be helpful to the network if there was a way for the node to ping the satellites to indicate that it’s offline, so that no traffic is even attempted to be sent to it, and then ping again when it comes online, to indicate that it’s open for business?
From a node operator perspective, maybe this would be a reduction in the “offline penalty” compared to a truly unplanned outage.
only really matters if it would benefit the network or the clients.
It won’t make any difference. Network already handles offline nodes regardless of reasons with 4 hour (I think) grace period. From customer perspective it’s also irrelevant — data is sent to more nodes than necessary, so the corner case where node “just went online” is covered and is transparent.
I’d go as far as claiming that it’s better for the network to not rely on additional signals like these, because unplanned downtimes are worse, and yet they have to be handled gracefully. Hence, no point in optimizing planned ones — the same mechanism can and should be used. Moreover, planned downtimes can easily turn into unplanned ones — so it’s best the network assumes the worst from the beginning.
Either a node is making data available… or it’s not and the repair system may need to get involved to make it available. A flag to indicate a node may come back online, maybe, probably, hopefully at some unknown time in the future does nobody any good.
And not getting paid anything if you aren’t doing anything… is the default. That’s not an “offline penalty”
As far as I can see the first significant penalty for nodes (suspension) kicks in after being 12 days offline. This should be more than enough for maintainance. So no need to change things from my point of view as a SNO.
If the satellite detects the node as offline, it’s not offered to the customers. After 5h all pieces on such an offline node marked as unhealthy and can be recovered to other online nodes.
Your idea makes no sense, because the node can be offline up to 12 days, then it would be suspended and you would have a week to bring it online back. After 30 days offline it will be disqualified. Plenty of time to make any maintenance, and it will not be free of penalties to incentive to keep the network healthy.
It will not, because of outlined behavior above. The offline node is still the offline node. If there would be enforcement of the requirement of 5h downtime per month, your node would be already disqualified. Of course, the offline node which may be return online is better than node, which losing or corrupting data, but well.
Your idea also a duplicate: Search results for 'maintenance #ideas-and-suggestions' - Storj Community Forum (official)
