DPE-4658 global-primary on endpoint

[paulomach]

Issue

1. Slow status update on switchover
2. Endpoint being empty breaks related router
3. cluster name clash workaround use meaningless hash

Solution

• propagating switchover status change across cluster
• use global cluster set primary on standby endpoint
• use model_uuid to workaround cluster name clash (instead of generated uuid)

[carlcsaposs-canonical]

Endpoint being empty breaks related router

IMO, this shouldn't be fixed with

use global cluster set primary on standby endpoint

because router will still fail on primary switchover if previous primary goes offline (e.g. region failure, network cut)—and this only hides the issue

but does this improve UX for consuming applications when async enabled?

[paulomach]

Endpoint being empty breaks related router

IMO, this shouldn't be fixed with

use global cluster set primary on standby endpoint

because router will still fail on primary switchover if previous primary goes offline (e.g. region failure, network cut)—and this only hides the issue

but does this improve UX for consuming applications when async enabled?

The main point here is to not disrupt a running service due a switchover (failovers are a complete different beast). Fact of matter is that we still don't have the full router on async history defined by Mohamed. Meanwhile this indeed avoids unnecessary disruption for end user.

[carlcsaposs-canonical]

The main point here is to not disrupt a running service due a switchover (failovers are a complete different beast). Fact of matter is that we still don't have the full router on async history defined by Mohamed. Meanwhile this indeed avoids unnecessary disruption for end user.

is this solution helpful for non-router apps? or router only?

[paulomach]

is this solution helpful for non-router apps? or router only?

Specially helpful for non-router, but we always recommend router for async setup.