Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Fetching contributors…

Cannot retrieve contributors at this time

128 lines (108 sloc) 4.387 kB
2.4.1
-----
- Only rewrite journals for empty queues or when in read-behind
2.4.0
-----
- Server status: up, read-only, quiescent
- Support for service discovery via ZooKeeper
- Kestrel now ignores directories in its queue path.
- Fix load test scripts to use correct JAR name [Bryan English]
- Scala 2.9.2, finagle 5.3.x
2.3.4
-----
- modify configuration to allow inheritance by fanouts to masters, masters
to default.
- log client description when queues are created, deleted or flushed
2.3.3
-----
- track queue item add time and compute duration on request
- change kestrel.sh to use old PID file so that daemon-based installs can
be halted
2.3.2
-----
release: 23 Aug 2012
- support gets in memcache protocol [David Pershouse]
- add transactions, canceled_transactions stats per-queue
- kcluster updates
- ksniff tcpdump helper
- aliased queue tests [Radu Voicilas]
- removed daemon from kestrel.sh to resolve SIGHUP problems
2.3.1
-----
release: 16 July 2012
- Replaces 2.3.0
- Introduce queue aliases, see docs/guide.md
- Remove all queue-related stats when a queue is deleted [Cory G Watson]
- Fanout queues are not tracked until they are successfully created (old
behavior caused never ending set errors if a fanout queue could not be
created due to filesystem name limitations, for example).
- Non-existent queues are no longer created
- when transactions are confirmed or aborted
- on flush
- on delete
- when stats for the queue are retrieved (e.g. STATS <queue>)
2.2.0
-----
release: 11 May 2012
- Thrift interface. See src/main/thrift/kestrel.thrift. [Jason Jackson]
- Kestrel now uses the Finagle library for networking.
- Allow max items to be specified for memcache protocol monitor command
- maxExpireSweep configuration to limit the number of expired items removed
from a queue per sweep [Hitoshi Asai]
- maxQueueAge configuration auto deletes empty queues that have existed
longer than the specified time [Cory G Watson]
- New statistics:
queue_creates, queue_deletes, queue_expires (global) [Cory G Watson]
cmd_monitor, cmd_monitor_get (global)
total_flushes, create_time (per-queue) [Cory G Watson]
- Better error reporting on startup failure
- Scala 2.9.1, sbt11, sbt-package-dist 1.0.5
2.1.5
-----
release: 12 January 2012
- Only sync to disk when there's data to be written. [Stephan Zuercher]
- Track latency measurements better, and track the percentage of java heap
reserved for in-memory queues. [Stephan Zuercher]
- Fix the startup script to work in "dash". [Matt Parlane]
- Send the correct reponse back for the memcached protocol for delete. [Matt
Erkkila]
2.1.4
-----
release: 21 November 2011
- Separate timers for journal fsync operations from those used for request
timeouts and queue expiry
- Additional statistics include set/get latency in microseconds and actual
request timeout length in milliseconds
- Upgrade to latest Netty, Ostrich and Util libraries (fixes for Netty timer
bugs)
- Creating a queue with an illegal name causes an error
2.1.3
-----
release: 13 October 2011
- Fix a bug where timers could stick around even after a client disconnected
or received an object.
- Add 2 missing server stats to the ostrich port.
2.1.2
-----
release: 29 September 2011
- Fix bug 71: connection counts could go negative. [Jeff Clites]
- Allow a queue name parameter to "dump_stats". [Jeff Clites]
- Fix potential file descriptor leak when rotating journals. [Ross Cohen]
2.1.1
-----
release: 23 September 2011
- Remove the extra linefeed from memcache "version" and "dump_stats"
repsonses. [prcm]
- Reduce the number of Promise objects created when a client isn't waiting
for one.
- Use a separate background thread for expiring queue items.
- Several improvements to the load tests. [Eric Lindvall, Jonathan Reichhold]
- Add "version" support to the text protocol. [Cory G Watson]
- Report latency of gets (hits vs misses) and sets.
- Fix a deadlock with transactions that abort and hand off the aborted item
to a waiting client that has also disconnected. [Jeff Clites]
- Don't fsync if nothing has been written to the journal since last time.
[Jeff Clites]
- Be more careful about recovering from corrupt/truncated journals, and exit
cleanly if something goes awry during startup. [Jeff Clites]
- Delete counters & gauges for queues that have been deleted. [Cory G Watson]
Jump to Line
Something went wrong with that request. Please try again.