As discussed in #10220, MCP and ADS protocols allow endpoints ( and the equivalent ServiceEntry) to be sent incrementally, with the server sending updates with only the clusters(services) that have changed.
The pilot side changes are relatively small - I can take care of them ( unless Nino is willing to do them to get rid of the diff for CF ).
The main gap is having MCP server ( galley, CF or a small standalone/dedicated server dealing with endpoints only ) that sends only the ServiceEntries that changed - and a lot of testing and product-ionization. If the tests go well we can even use this for the primary cluster - but I think it's safer to have it as an alternative for MC.
Let's try to sync up - I think this is doable in the 1.1 timeframe.
As discussed in #10220, MCP and ADS protocols allow endpoints ( and the equivalent ServiceEntry) to be sent incrementally, with the server sending updates with only the clusters(services) that have changed.
The pilot side changes are relatively small - I can take care of them ( unless Nino is willing to do them to get rid of the diff for CF ).
The main gap is having MCP server ( galley, CF or a small standalone/dedicated server dealing with endpoints only ) that sends only the ServiceEntries that changed - and a lot of testing and product-ionization. If the tests go well we can even use this for the primary cluster - but I think it's safer to have it as an alternative for MC.
Let's try to sync up - I think this is doable in the 1.1 timeframe.