Why does it matter if gracefully exit or not?

In the docs it says:

The withholding model is designed to incentivize and reward both-long term reliable Storage Nodes as well as Nodes that, when they do choose to leave the network, exit in a way that is least damaging to the network.

but why does it even matter? I imagine that the pieces have to move to a new node, no matter if the node leaves the network gracefully or not. Why and where is a difference between a node leaving gracefully or not?

I believe only two things:

  • allow the repair workers to use your pieces to repair segments if it’s required;
  • get your held amount back.

The last one I believe would be matter after these latest tests.

But why would that be necessary? With the expansion factor of 2.3?

I can imagine it being less compute intensive, since you just copy the piece instead of recalculating it like in a RAID. Is that a reason?

yes, it’s could be necessary. Some segments may require the repair right away. However, it’s a free participation. You may leave it abruptly, it’s up on you.

we never copy pieces. We can regenerate them using the minimum required pieces for the selected Reed Solomon settings. Each piece is unique accordingly that math, no replication/clone/duplication is needed, only the minimum required amount of unique pieces of the same segment.

Maybe this is again a language barrier between us.
I read your answers like a snarky “if you don’t like it you can leave!”

Let me try again. I am not debating about leaving the network or not. I also don’t care about payouts! I just wanna know what the difference for the network is, if a node leaves gracefully or not.

Since we have some kind of built in parity, there should no difference between the two, right?

Or maybe we can use a different analogy. If I have a 8 wide RAIDZ2, the RAIDZ is like the storj network and a single HDD like a node.
Leaving the network gracefully is like saying ZFS I wanna replace one disk.
Leaving the network not gracefully is like yanking one disk out and put in a new one.

There should be no difference (except that the first one is a little bit more secure since in case of a checksum error that drive could also be used).

Oh, I’m sorry. No, I do not meant to sounds like that, clearly.

The differences between an abruptly leaving the network and perform a new version of a GE are:

  • you would get a held amount back;
  • you probably will help the network, if your node is keeping required pieces to repair a segment;
  • just be a good SNO, who want to help the network, even if the exiting node would not help the repair workers, because everything is fine already.

Cheers mate.

So leaving out the money which I don’t care about, the only difference is the stability of the network. And an incentive not suddenly just pull the plug. Got it.

But why would that be even needed? With an expansion factor of 2.3, there should be more than enough empty nodes that can take over and still are able to pull the data from the remaining nodes.

Expansion factor is good, until the repair threshold is reached. It’s possible, that the repair workers are failed to do their job before it’s too late. And it’s possible, that your node is keeping the last mandatory piece to repair the segment in time.
We do not want to play with possibilities, so have your exiting node for the next 30 days is a good backup, if you do not mind.
However, it should survive accordingly stat… But it’s always good to increase chances, isn’t it?

2 Likes