Glusterfs: Asynchronous, Distributed distant-replication using journals in glusterfs.
Submitted by Ajeet Jha (@ziiin) on Wednesday, 12 March 2014
Disaster recovery and filesystem synchronization between geographically distinct locations are still ambitious features in filesystem world. "Geo-Replication", a feature of glusterfs is an influencive solution to it.
If a node storing and maintaining your data is down, you have a way to get your data back from one of the "near-replicated" nodes. But what if your complete replicated data setup at that location is down, damn!! you are screwed.
Geo-replication can recover your data even after such scenarios.
This was indeed an impressive feature of data recovery post disaster. Now, if you want a maintenance free sync of your filesystem data to a location way far from your current location.
Geo-rep provides this unique ability within glusterfs, just ask for sync and it will be done :)
An additional feature used by geo-replication is glusterfs journals (changelogs). These are granular journals, that record file-operations which occur in that time-frame (configurable).
P53ud0 C0d312 at RedHat and always trying to answer the question, "Can we do better?".
Interested in distributed filesystems, especially distant replication.
Hugely motivated from and supported by:
- Venky Shankar (Engineer at redhat and presenter at Rootconf-2013).
- Aravinda Vishwanathapura Krishna Murthy (http://aravindavk.in/)
- Kotresh Hiremath Ravishankar (Engineer at RedHat. Fascinated by system side programming and very much interested in filesystems).