It’s been another three weeks and v0.47 is ready. The highlights include:
- mon: admin tools to control unwieldy clusters (temporarily block osd boots, failures, etc.)
- osd: reduced memory footprint for peering/thrashing
- librbd: write-thru cache mode
- librbd: improved error handling
- osd: removal of ill-conceived ‘localized pg’ feature (those annoying PGs with ‘p’ in them)
- rest-bench: simple tool to benchmark radosgw (or S3) (based on ‘rados bench’ command)
In truth it wasn’t the most productive sprint because of the work that went into the launch of the web sites, the launch party, and the subsequent inebriation. However, the new RBD caching feature is looking very good at this point, and patches are working their way upstream in Qemu/KVM to enable it with the generic ‘cache=writethrough’ or ‘cache=writeback’ settings.
One other noteworthy item is that I generated a new PGP key to sign releases with. The key is now in ceph.git, and has been signed by my personal key. If you are installing debs from our repositories, you’ll want to add the new key to your APT keyring to avoid annoying security warnings.
For v0.48, we are working on a ceph-osd refactor to improve threading and performance, multi-monitor and OSD hotplugging support for upstart and Chef, improvements to the OSD and monitor bootstrapping to make that possible, and RBD groundwork for the much-anticipated layering feature.
You can get v0.47 from the usual places:
- Git at git://github.com/ceph/ceph.git
- Tarball at http://ceph.newdream.net/download/ceph-0.47.tar.gz
- For Debian/Ubuntu packages, see http://ceph.newdream.net/docs/master/install/debian
a new website + commercial support, but does anyone have an estimate of when it will be production quality?
The definition of “production quality” varies depending on who you ask; I don’t like to treat it as a binary term.
At this point we support the RADOS object store, radosgw, and rbd because we think they are sufficiently stable that we can handle the support workload. There are several organizations running those parts of the system in production. Others wouldn’t dream of doing so at this stage.
We can tell you how we test, and what we support, but in the end it’s your judgement that matters.
and the ‘file system with POSIX semantics’?
We will be shifting focus to the file system later this year, but I can’t give you a meaningful estimate on “production-ready” at this point.