Here's how Subversion Edge replication works, why replication could be useful in your setup, and what to consider when you plan to set up a replica.

Why replicate?

Typically, you would deploy a replica for one of these reasons:
Projects in remote locations with lower bandwidth or higher latency want the performance of a "local" server.
Your company has a number of developers clustered together at a remote location. When you install a replica inside the LAN of these developers, they can greatly improve their Subversion performance and keep a lot of their traffic off the WAN. In this scenario, you would probably want a replica in each such location. Keep in mind that a replica can only be a proxy for one master — so if your company has more than one Subversion master server, you may need more than one replica at each location.
You want to reduce the load on the master server.
For example, continuous integration tools can place a lot of load on the server and moving that load to a separate server can increase the response time for other users. In this scenario you probably only need to add one replica; you'd add it as close to the master as possible so that synchronization is quick. Of course the previous point can be a factor here. If the continuous integration server is at remote location, then you would want to put a replica near the continuous integration server.

Rules for using a replica

When you convert a Subversion Edge server to replica mode, you provide the TeamForge username and password to use for the replica. The replica uses these credentials when it replicates Subversion content. This user must be added to the TeamForge project(s) and given permissions to the repositories being replicated. Those permissions also control what parts of the repository will be replicated. So if you have folders that should not ever live on remote servers, you can set up path-based permissions and that content will never be replicated to the server. If you forget to set up permissions, the replication will fail. However, there's no real harm done, and once you fix the permissions, you can do it again.

The replica user can be a normal user account — it does not have to be an Admin account. If the replica is set up and maintained by an Operations team, they might want to just use an Admin account so that project teams do not have to worry about adding the user to the project or setting up permissions.

Permissions for end users accessing those repositories will follow the normal TeamForge rules.


All communication originates from the replica. TeamForge never contacts or pushes to the replica — the replica initiates this. When TeamForge wants the replica to do something such as synch a new repository or synch a new revision for an existing replicated repository, it queues an event for that replica. When a new commit comes in to TeamForge and a new commit object is created, the TeamForge application server queues up an event for each replica that has this repository. If the commit is for a repository with no replicas, then nothing happens.

The replica polls the TeamForge application server for its events. TeamForge site administrators can configure the polling interval for each replica individually. Typically the interval would be around 60 seconds. The longer the polling interval, the longer it will take for a commit to eventually reach a replica. At the same time, the longer the interval, the lesser the load placed on the application server by the polling mechanism.

The replica receives all of the queued events since the last time it polled, and splits these events into two groups: TeamForge site administrators can control how many simultaneous jobs of each type run concurrently. Typically, the number for the first group would be low, such as 1 or 2. For the second group, it would be larger. Obviously the maximum simultaneous svnsyncs will have an impact on the Subversion master and the replica itself and must be considered. For more information on replica settings and how to configure them, see Edit replica settings.