Backups for SNOs

Node Operators
1

I’m reviewing best practices for storage node operators.

In this thread, the consensus seems to be that there is no point in backing up hosted content, so SNOs are advised not to do any backups.

Is this still correct?

2

Currently, the situation is that there is little penalty for downtime.

I currently use RAID to prevent against disk failure, because I care about being an aged node with little escrow withholdings.

I don’t see a need for live failover unless the penalty is brought back for 5hr downtime during a month DQs a node. Then, it might be worth looking into some type of hot failover. But probably won’t be financially a great idea unless you already have the equipment.

3

The problem is that it is not possible to have an actual backup of the node. The satellite has a list of all the pieces the node should have and if you restore a backup that’s an hour old, the node will be missing some of those pieces, leading to failed audits and disqualification.

There is no way to tell the satellite that “this node has just been restored from backup and does not have any pieces uploaded to it after date xyz”

So, only RAID is useful here, or HA. Still if the database gets corrupted and is not possible to repair, there is no way of restoring it from backup.

4 Likes
4

Hey @pentium100, this is useful info.

Seems to me there should be some kind of “suspend node temporarily” functionality, which SNOs could invoke intentionally (ranging from “Hey, the electrician is coming to my house today and I won’t have power for 8 hours” to “I’ll restore from backup”) and which would allow the satellite to do whatever necessary to resync after resume is complete.

I understand that would make things more complicated, but nodes will go down occasionally in any case and it would, IMHO, be better to do this in an “announced” fashion whenever possible than not knowing what the reason for the node outage is. Same reason sites announce maintenance intervals rather than simply not responding.

V.next feature perhaps :slight_smile:

2 Likes
5

Also, SNOs could be given a fixed number of hours each month for “maintenance downtime” in addition to the 5h of allowed “unplanned downtime”.