On 3/07/2014 11:50 PM, Richard Newman wrote: >> So you do have something ready for large scale or at least redundancy. My >> main concern is about "if my server crash ... how my users will sync". > > Sync isn't designed to require high availability. Clients will retry quietly > if their storage server is unavailable. It doesn't even require durability: > because clients have a full copy (and more besides) of all the data on the > server, if a DB dies we can just punt them to a different storage server and > the client will reupload.
I'll second this: do not try to make the sync storage db redundant. If it dies, the correct approach is to stand up a new node and let the clients re-upload a fresh copy of their data. If you want to provide more of a "backup" service than a "sync" service, you can take periodic snapshots of the storage db and restore from one of them in the event of failure. But the sync client and protocol are not really designed for this. Ryan _______________________________________________ Sync-dev mailing list [email protected] https://mail.mozilla.org/listinfo/sync-dev
