Release Notes

We would like to thank everyone who contributed to this release, including but not limited to:

We would also like to acknowledge and extend a special thank you to Ceph veteran, Sandon Van Ness, who passed away unexpectedly on February 23, 2015. He was 29 years old.

Red Hat Ceph Storage v1.2.3 is the first release of Ceph Storage using the Red Hat build process. This version includes all the features of Inktank Ceph Enterprise v1.2.2 and the enhancements described in these release notes. Between v1.2.2 and v1.2.3, there were no major features added.

Enhancement to Ceph monitors include fixes to rare corner cases for Paxos, and new Paxos instrumentation. Ceph monitors also abort start-up if the monitor’s data disk is too full. Ceph monitors also have improved osdmap trimming.

Enhancements to Ceph OSDs include adding new erasure-code regression tests, adding use-after-free code in cache tiering code, handling EC + primary affinity misdirected op checks properly, and mounting XFS with inode64 by default.

Erasure-coding and cache tiering are tech previews only and are not supported for production clusters.

Ceph OSD journals have enhancements to direct-io shutdown, journal dump, snap trimming performance, and snapdir handling on cache eviction.

Peering enhancements include clearing rollback PG metadata on PG deletion, addressing occasional stalls during peering or activation, and displaying past_interval values correctly.

Backfill enhancements include making backfill reservation denial handling more robust, fixing full OSD checks during backfill, and removing bad backfill assertions for mixed-version clusters.

Additional enhancements include cleanly aborting set-alloc-hint operations during upgrade; cancelling callback on blacklisted watchers; not aborting deep scrub if hinfo is missing; parsing ioprio properly; processing admin socket dump_ops_in_flight properly; handling map gaps in map advance code; handling undefined CRUSH results in interval check; including shard in JSON dump of ghobject; and, fixing invalid memory reference in log trimming.

Block device enhancements are to librbd, and include capping memory utilization for read requests; preventing closure of a closed parent image on failure; ensuring diff tests run properly; and, protecting list_children from invalid pools.

The Ceph kernel block device is only available for RHEL 7 and beyond. It is not available or supported on RHEL 6 and earlier versions of RHEL.

Gateway enhancements include packaging improvements; ensuring sysvinit scripts run the Ceph Object Gateway under a systemd context as needed; and, using standard udev rules for RHEL7. The gateway now adds .log messages to the default log path and separates civetweb logs from the rgw log. There is also a fix for civetweb URL decoding.

The gateway uses mod-proxy-fcgi instead of mod-fastcgi, and includes some clean up for the FastCGI request context.

The gateway now sets a length for Keystone token validation, and improves subuser creation for Swift. There is also a fix for the object tail test.

The gateway has a few improvements for REST calls; it will no longer try to authenticate CORS preflight requests; calculates hash codes properly during PUT operations; and, it converts header underscores to dashes.

On copy operations, the gateway will now copy object data even if the copy target is in a different pool from the copy source.