Add force sync REPLICA with MAIN

[antoniofilipovic]

This PR enables the cluster to continue executing if the MAIN is behind REPLICA. It can happen in the scenario as in the test case, or when one of the tenants on other REPLICA is behind MAIN's tenant. In both cases, the result is the same -> force sync of tenant with MAIN's tenant.

This is as agreed in user story for 2.15/2.16 -> Link here

[master < Task] PR

• Check, and update documentation if necessary
• Provide the full content or a guide for the final git message

To keep docs changelog up to date, one more thing to do:

• Write a release note here, including added/changed clauses
• Tag someone from docs team in the comments

[andrejtonev]

General question about the scheme; why do we care about epochs? Shouldn't everyone have the same history? If the epoch do not align, then we have a branching point and the two instances cannot continue as part of the same cluster.
I understand that we might decide to delete data in that case. But I still do not understand why we care about epoch even in this case.
We care about who has the highest timestamp.
In case we have a branching point we might care about minimizing the differences. Not sure about this either. We should prioritize deleting the least amount of data. Timestamps don't exactly correlate with "amount of data" but should be enough at this point.

[andrejtonev]

First pass. A few comments.

[as51340]

I will leave this one to Andreja to handle it, high-level looks good to me

[antoniofilipovic]

General question about the scheme; why do we care about epochs? Shouldn't everyone have the same history? If the epoch do not align, then we have a branching point and the two instances cannot continue as part of the same cluster. I understand that we might decide to delete data in that case. But I still do not understand why we care about epoch even in this case. We care about who has the highest timestamp. In case we have a branching point we might care about minimizing the differences. Not sure about this either. We should prioritize deleting the least amount of data. Timestamps don't exactly correlate with "amount of data" but should be enough at this point.

We synced about this. So the issue was that the Notion page was out of sync. I don't deal with epochs here.

[andrejtonev]

Looks good!

[antoniofilipovic]

@kgolubic tagging you here:
RN: In case REPLICA diverges from MAIN, MAIN will force reset data on REPLICA instance

[antoniofilipovic]

Docs here:
memgraph/documentation#630