Replication Advice When A Node Has Been Absent For A While


#1

Hello:

I run a cluster of a couple of machines used for archives: users uploaded data onto one machine (“Master”), and this gradually synced to the other (“Slave”). Both machines ran in Send & Receive mode.

Unfortunately the Master suffered a hardware issue before the sync completed (there was a lot of data), so I had to pause the sync and manually transfer outstanding data across to the Slave, and then swap roles - the Slave was now being updated by users.

I’ve since resolved the issues with Master, and need to get it back in sync with Slave - but I’m nervous about just setting the sync just running again, in case the Master starts to undo (wanted) changes on the Slave.

What would you advise I do, prioritising safety of files on the Slave above all else? I’m considering two options:

  • Remove the Folder from Syncthing on Master, actually delete the folder from the Master filesystem, then re-add the Folder into Syncthing on the Master. Will this result in Slave pushing all files onto Master, and prevent any deletions from pushing onto Slave?
  • Set Slave Folder to Send Only and Master Folder to Receive Only, until they are both fully in sync. Presumably this will allow it to take advantage of the already-synced data, without needing to re-transfer it. At a later point, I can set them both back to Send & Receive if necessary (as there might be data uploaded onto both units in the future).

Any thoughts?

Thanks,

Pants.


(Jakob Borg) #2

I would set the out of date machine to receive only, let it scan and reconnect to the cluster. It’ll suck in any new updates. It may think it has some data that is newer than everyone else, in which case you will get a Revert button. Use that, and it’ll get back in sync with everyone else.


#3

Great - thanks Jakob!