Hi everyone, this article (http://blog.dovecot.org/2012/02/dovecot-clustering-with-dsync-based.html) describes the situation for director/NFS-based clusters and SSH-based clusters. Right now I'm running the first setup with one proxy/director and two backends (all 2.2.x) mounting an NFS-share but I'd like to get rid of NFS (it's slow) and, if one backend is down, the proxy still sends the user to that backend. The article says: "High-availability non-NFS setup One possibility is to use Dovecot proxies, which know which servers are down. Instead of directing users to those servers, it would direct them to replica servers." My question is, if I can still use my proxy/direcot setup but every backend having it's local (or iSCSI-backed) ext4 storage and not using the SSH-based replication method? Regards Patrick
Hi Patrick, You can use the proxy in combination with a database to return a fixed backend host per user, but be aware of the fact that if one host goes down, all mailboxes on this host will also be down (if there is no replica). The director uses a hash function instead and can only be used when all backend nodes can serve the same set of users. The director uses the proxy feature to forward the connection to the backend host, but the backend is selected by the hash function depending on the username. In short: use the proxy not the director, but keep copies/replica of your mailboxes somewhere in case one host goes down. Regards Daniel
Apparently Analagous Threads
- Dovecot clustering with dsync-based replication
- add geo-replication "passive" node after node replacement
- add geo-replication "passive" node after node replacement
- Syntax for creating arbiter volumes in gluster 4.0
- Initial replication halts with "The handle is invalid." (msDS-NC-Replica-Locations corrupted?)