Apigee Edge Synchronization Issue Between Two Data Centers

[OsamaGamal]

I have two data centers (DCs) in Apigee Edge: one is running, and the other is stopped. When I deploy an API, it gets deployed to the running data center. However, when I start the stopped data center, I notice that the API has not been deployed there.

Is there a way to synchronize the API deployment between the two data centers, even if one of them is stopped? Specifically, I need a solution that ensures synchronization happens automatically once the stopped data center is started.

[paul-wright]

Hi OsamaGamal,

Welcome to the community!

Please confirm a couple of items for me:

1) It sounds like this is a version of Apigee Edge for Private Cloud, what version of OPDK is being used?

2) What the 2nd data center is "stopped", does that mean all of the Apigee services are stopped (not running) or does that refer to just the traffic flow being stopped?

3) What topology is being used? https://docs.apigee.com/private-cloud/v4.52.00/installation-topologies
(please adjust the version Apigee appropriately in the link above) 

Thanks.

[OsamaGamal]

Hello Paul,

The current version is 4.52.02-0.0.1139.

By "stopped," I mean that I’ve shut down all machines and Apigee services in the second data center.

My setup:

Apigee installation across Multiple Data Centers

6 nodes per data center

I’d also like to mention an issue I encountered:

If I deploy a new API while both data centers are running, and then shut down one of the data centers, I’m unable to deploy a new revision of that same API on the remaining active data center.

To deploy a new revision, both data centers must be up and running

[dknezic]

Apigee multi-region isn't designed to run with a DC shutdown.. 

[paul-wright]

Hi OsamaGamal,

As Dane mentioned above, Apigee OPDK is not designed to have a regional data center shut down for extended periods of time. OPDK is designed to be active / active having multiple data stores actively replicating data between data centers as configurations and application data such as authorization tokens are created or updated.

You can see in our upgrade scripts that we generally only update one component at a time to ensure both the uptime of the system and the data replication between data centers.

If, for some reason, a data center is shut down for more than about 90 minutes, a rebuild of the cassandra nodes in the downed data center will be required to bring the dc back online.