Upgrades from shared to dedicated clusters should only make the cluster available after data has been restored
We have recently upgraded from a shared tier to a dedicated cluster. As document, this process involved a downtime. However, what was not document is that the cluster becomes reachable for clients again before the data has been fully restored. That way, clients can invoke operations on the cluster, that may even interfere with the mongorestore process. I would propose to make the cluster available again to public clients only after all data has been restored to avoid the chance for such race conditions.
Johannes
shared this idea
-
Hi Johannes,
Thanks for filing this: We are working on a next-generation platform enhancement that will offer a different architecture and way of scaling in a more liquid manner without downtime or the other rough edges you've mentioned here. This is a major undertaking but stay tuned for next year: I wish we had something for you sooner.
-Andrew
