I'm looking for something that can replicate file systems between multiple branch locations. Our file servers currently run CentOS 7.

I feel like there has to be a better solution that what we're using now. Currently we have 2 branches, and I'm using lsyncd with csync2 bidirectionally to keep the filesystems in sync. We're adding a 3rd branch soon, and while the lsyncd/csync2 combination can do that, it has already shown that it can be fragile with only 2 locations, and I'm afraid adding a third will make it worse. The branch offices are connected by a VPN over DSL/Cable, so anything that can't survive a high latency connection won't work (eg: GlusterFS)

Most of the replicated filesystems I've encountered are designed so that the replica is meant as a failover, so the replica can't be used unless it's the master. I need the replicas to be usable for read/write at all times.

It seems like this would be a common scenario in business, with multiple branch offices, but I can't seem to find anything that looks like it will work the way I want.



  • Can you elaborate on your current setup? How lsyncd and csync2 interact? – shodanshok Apr 27 '16 at 5:55
  • I'm using an updated version of the method described here: scribd.com/doc/182910221/… It's been updated only to make it compatible with recent versions of Lua. The reason I'm worried about expanding this to 3 locations, is that in the above method, each location updates to another single location in a ring, which would cause excessive delays should one of the locations lose connectivity. – Analytical360 Apr 27 '16 at 18:52
  • Oh, I see, In the past months, I evaluated similar methods to synch two fileservers on different offices, but with poor results. As my customer wanted near-synchronous replication, I ended writing my own synchronization software. However, it is limited to two branches. Have you evaluated a star topology replication using Unison (see here: cis.upenn.edu/~bcpierce/unison/download/releases/stable/…)? I my case, the large dataset (500+ GB, ~700.000 files) did not play so well with Unison or csync2, but maybe your dataset is smaller... – shodanshok Apr 28 '16 at 21:55
  • I think we decided on XtreemFS. We're hopefully implementing this tonight, so I'll update on how that works out for us. – Analytical360 May 5 '16 at 1:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.