Connector Improvement: Allow for updating or creating a new slot without requiring a full resync
PlannedWhen we upgrade a database or perform other types of maintenance it would be nice to be able to update the slot name or create a new slow without having to perform a full resync of the data.
We are able to guarantee that no app data is changed during the window when we are performing DB maintenance and it would be much faster to avoid the resync and just point the connector at a new slot and have begin reading. Currently we receive an LSN error if we just Unpause the connector, and have not found any way around this.
-
Official comment
Hi Garrett,
Thanks for sharing your request! We're currently investigating how to better support such scenarios. So it is on the roadmap, but not under active development yet.
Can you please share a little more about how you perform the maintenance? Do you use blue/green deployments of some kind? Do you stop application workloads for the duration of the switchover window? The more detail you can provide to us the better, as it will inform how we design the feature.
Regards,
Val Kulichenko, Fivetran Product Team -
+1 from Playlist. We upgrade Aurora PostgreSQL with AWS Blue/Green, which requires dropping the Fivetran replication slots before the deployment and recreating them afterwards. Many of our tables use History mode, so a historical re-sync is not a good option for us.
Today the path is the Support-assisted LSN cursor update. It works well, but it ties our maintenance window to Support availability. Being able to point a connection at a recreated slot (same name, new confirmed_flush_lsn) without a re-sync would make this a self-service, scriptable step. Thank you for considering it!
-
Hi Augusto,
Thanks for expressing the interest in the feature. We're actively working on better support for blue/green deployment, stay tuned!
Thanks,
Val Kulichenko, Fivetran Product Team
Please sign in to leave a comment.
Comments
3 comments