Open esune opened 11 months ago
Having the geographic redundancy between Kamloops and Calgary datacenters that migrating to Gold provides comes with the cost of added complexity. The applications need to be synchronized between the two instances, same versions deployed, etc. Data also needs to be synchronized (it's not automatic and the instances don't use the same data source), and our services are far more sensitive to synchronization than most.
For example, for an issuer service it would be critical to ensure the synchronization of the secure storage (wallet) across the two instances on both fail-over and fail-back. It's very possible for the issuer's secure storage to get out of sync with the ledger or it's clients (issuing and revocation) when it's using one secure storage instance one moment and a different one the next. You can see these issues in rocket.chat as it fails from one cluster to the other. The data is not synchronized in real time and you see massages come and go as rocket.chat fails-over and fails-back.
For an application like OrgBook the volume of data becomes and issue when it comes to synchronization.
So the question is really, does (or when does) the need for application availability out-way the effort required to overcome the added complexity and issues we'd have with application and data synchronization?
Thank you @WadeBarnes , these are really great points and good background information - I was not thinking of the added complexity we would be taking on.
Current namespaces are in OCP Silver, do we need to migrate to Gold or are we good (for now)?