Ceph status

[heathen]

Hello!

I have a project which needs a storage solution with strong redundancy, and not only with drives/file system fault tolerance, but also with full hardware failure tolerance (with two or more identical synchronized storage systems). Enterprise class solutions costs too much, so I'm looking for open source solutions.

On the drive/file system level I chose ZFS. But on the network/whole storage level I can't make the right decision: there are no working solutions on FreeBSD. HAST can't be used on production and it looks abandoned.

I read about Ceph, but I find nothing about its current status on FreeBSD. Does it work currently? How stable it is and is it possible to use it in production?

Thanks!

 

[gregsfortytwo]

The main body of upstream Ceph developers (employed by Inktank, including me) don't do any work or testing on FreeBSD. It's pretty generic POSIX code, though, and somebody got it building and working on FreeBSD several months ago — this work went upstream — so it shouldn't be difficult to get building even if we did something to break it.
That said, while the block and object storage systems are stable and supported I wouldn't recommend the filesystem for production work yet. It needs to go through more stabilization and QA first.

 

[mix_room]

HAST can't be used on production and it looks abandoned.

I wouldn't say that HAST looks abandoned. pjd replies to mailing list questions, as in http://lists.freebsd.org/pipermail/freebsd-fs/2013-January/016316.html

Give HAST a try http://www.freebsd.org/doc/handbook/disks-hast.html, it does have a handbook chapter, which is quite a good indication that it is, at least weakly, supported.

 

[phoenix]

HAST can't be used on production and it looks abandoned.

HAST most definitely is not abandoned. There's multiple commits to -HEAD for HAST each month, sometimes multiple per week. Haven't kept up on HAST commits to -STABLE, but things are usually MFC'd within a month of hitting -HEAD.

Haven't personally used it since it originally hit 7.x, but have read about people using it successfully.

 

[estrabd]

I am interested in ceph on FreeBSD, and it's important to realize that there are basically three ways to interface the storage cluster. The POSIX compliance is for the CephFS (ls, cd, etc). There is however a HTTP interface that talks both Amazon S3 and OpenStack Swift. The third is for a block device. I would imagine that while all three would be of interest, it would be easiest to investigate the gateway interface since it would presumably not require any low level FS stuff (just need a web server that support FastCGI).