Distributed sync
Hi guys, as suggested by @MikeT I wanted to move the end of this https://discussions.agilebits.com/discussion/comment/284808#Comment_284808 discussion to teams
My company would be very interested in using the on-prem version of this, and possibly the hosted, but I have some questions around how the sync is set up to handle netsplits and byzantine failures as well as access control and auditing. Is there a tech overview of what algorithms and data structures your server is using to handle coordinated replication and tests against say Jepsen to handle failure modes?
1Password Version: Not Provided
Extension Version: Not Provided
OS Version: Not Provided
Sync Type: Not Provided
Comments
-
@Nimish Telang In the past 10 years we re-implemented sync at least 5 times. We were never happy with it, even though CloudKit-based implementation was pretty good.
This is the reason we changed our approach in 1Password for Teams. The server is the primary source of all data and the clients simply keep an offline cache of it. In short, we tried to keep things as simple as possible and it turned to be the best solution we ever made. We have the internal documentation on this but we don't plan to make it public at the moment.
I hope to publish an overview of our server-side infrastructure at some point. It might put on-premise deployment out of reach for many teams as just cost of running all Amazon Web Services components is going to be in thousands $$ per month.
0 -
@roustem It would be nice to know, but i'd be surprised if self-hosting would be that expensive, assuming the technology scales up and down.
Anyway, definitely interested in how you get sync to work and not accidentally delete secrets or leave the db in a failed state
0 -
Than you, @Nimish Telang!
I'll try to get the post finished, hopefully soon.
0
