Running Wildduck instances on multiple servers.

Hi *,

I want to run two Wildduck instances on two different physical servers. The first one should be the main mail server and the second one should serve as a backup mail server, accepting mail in case the first mail server is down. Hence I want both servers to have identical data. Does anyone know how to do this, especially regarding mongodb and redis? @andris9 do you have any suggestions on this one? Any information is appreciated.

What I already looked into

All mails are stored in the mongodb database, therefor the database has to be synchronized between both servers. I’ve figured out, that mongodb has a feature called replication set that does exactly what I need.

I do not know what the corresponding concept is called for redis though or if I need to replicate the information in redis at all (haven’t looked into the code yet).

[andris9]

With 2 servers you could run Redis in a “normal” 1 master - 1 slave configuration without Sentinel and also 2 node MongoDB replica set (with or without arbiter). If one of these servers goes down then:

1. If the invalid server was running Redis Master, then configure Redis Slave to be new master (by running REPLICAOF NO ONE command in the former slave instance). Also modify WildDuck config files to use the new server as Redis host.
2. Unless you used MongoDB arbiter setup and the server with arbiter is still up then MongoDB goes down even if the remaining server was primary as it does not have enough quorum votes. So you would have to modify the replica set by removing the invalid instance from cluster configuration using rs.remove() command. Once you have a replica set with a single server then you have enough votes and the instance could become primary again.

[rqelibari]

Thank you all. I think @andris9 did explain the possible solutions to this issue very thoroughly. I for myself will use a three server setup to have automatic failover.