I was reading the old IBM docs on the openafs web site about replication and they say you should only use the replication on volumes that are infrequently updated, and possibly only managed by administrators since you need to run a "release" operation to update the read-only mirrors.

Is there a way to replicate an OAFS volume that will allow one user to save (and close) a file on a replicated volume and be able to know with confidence that another user (using another machine) will be able to access that updated file?

I don't have an environment with a lot of rapidly changing files, but the files that do change (office files, images, etc.) need to be accessible from other locations once they are saved. Would I have to make sure those volumes are not replicated?

Last, but not least, I'd really like it if there was some way to use OAFS in a Linux-HA environment where the r/w volume fileserver goes down and a r/o fileserver for those volumes takes over as r/w (using STOMITH to prevent a split-brain from happening (two r/w fileservers that don't know about the other). Is that a possibility?

I really like the caching affects that OAFS gives you, but for a r/w HA environment, Linux with ext3, LVM, DRDB, ACLs (& NFS4 in the future) looks promising, but still doesn't give active fileserver replication (I saw somewhere using DRBD or ENBD with one r/w and one or more r/o GFS fileservers together would give active replication, and the possibility of switching one of the r/o to r/w looks promising).

In summary, is OpenAFS moving in the direction of active replication?

Thanks,

Mike
_______________________________________________
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info

Reply via email to