Kestrel

Kestrel is based on Blaine Cook's "starling" simple, distributed message queue, with added features and bulletproofing, as well as the scalability offered by actors and the JVM.

Each server handles a set of reliable, ordered message queues. When you put a cluster of these servers together, with no cross communication, and pick a server at random whenever you do a set or get, you end up with a reliable, loosely ordered message queue.

In many situations, loose ordering is sufficient. Dropping the requirement on cross communication makes it horizontally scale to infinity and beyond: no multicast, no clustering, no "elections", no coordination at all. No talking! Shhh!

For more information about what it is and how to use it, check out the included guide.

Kestrel has a mailing list here: [email protected] http://groups.google.com/group/kestrel-talk

Author's address: Robey Pointer <[email protected]>

Features

Kestrel is:

• fast

  It runs on the JVM so it can take advantage of the hard work people have put into java performance.

• small

  Currently about 2K lines of scala (including comments), because it relies on Apache Mina (a rough equivalent of Danger's ziggurat or Ruby's EventMachine) and actors -- and frankly because Scala is extremely expressive.

• durable

  Queues are stored in memory for speed, but logged into a journal on disk so that servers can be shutdown or moved without losing any data.

• reliable

  A client can ask to "tentatively" fetch an item from a queue, and if that client disconnects from kestrel before confirming ownership of the item, the item is handed to another client. In this way, crashing clients don't cause lost messages.

Anti-Features

Kestrel is not:

• strongly ordered

  While each queue is strongly ordered on each machine, a cluster will appear "loosely ordered" because clients pick a machine at random for each operation. The end result should be "mostly fair".

• transactional

  This is not a database. Item ownership is transferred with acknowledgement, but kestrel does not support grouping multiple operations into an atomic unit.

Building it

Kestrel requires java 6 and sbt 0.7.4. On OS X 10.5, you may have to hard-code an annoying JAVA_HOME to use java 6:

$ export JAVA_HOME=/System/Library/Frameworks/JavaVM.framework/Versions/1.6/Home

Building from source is easy:

$ sbt clean update package-dist

Scala libraries and dependencies will be downloaded from maven repositories the first time you do a build. The finished distribution will be in dist.

Running it

You can run kestrel by hand via:

$ java -jar ./dist/kestrel-VERSION/kestrel-VERSION.jar

To run in development mode (using development.conf instead of production.conf), add a stage variable:

$ java -Dstage=development -jar ./dist/kestrel-VERSION/kestrel-VERSION.jar

When running it as a server, a startup script is provided in dist/kestrel-VERSION/scripts/kestrel.sh. The script assumes you have daemon, a standard daemonizer for Linux, but also available here for all common unix platforms.

The created archive kestrel-VERSION.tar.bz2 can be expanded into a place like /usr/local (or wherever you like) and executed within its own folder as a self-contained package. All dependent jars are included, and the startup script loads things from relative paths.

The default configuration puts logfiles into /var/log/kestrel/ and queue journal files into /var/spool/kestrel/.

The startup script logs extensive GC information to a file named stdout in the log folder. If kestrel has problems starting up (before it can initialize logging), it will usually appear in error in the same folder.

Configuration

Queue configuration is described in detail in docs/guide.md (an operational guide). Scala docs for the config variables are here: http://robey.github.com/kestrel/doc/main/api/net/lag/kestrel/config/KestrelConfig.html

Performance

Several performance tests are included. To run them, first start up a kestrel instance locally.

$ sbt clean update package-dist
$ VERSION="2.1.0-SNAPSHOT"
$ java -server -Xmx1024m -Dstage=development -jar ./dist/kestrel-$VERSION/kestrel-$VERSION.jar

Put-many

This test just spams a kestrel server with "put" operations, to see how quickly it can absorb and journal them.

$ sbt "put-many --help"
usage: put-many [options]
    spam items into kestrel

options:
    -c CLIENTS
        use CLIENTS concurrent clients (default: 100)
    -n ITEMS
        put ITEMS items into the queue (default: 10000)
    -b BYTES
        put BYTES per queue item (default: 1024)

A sample run on a 2010 MacBook Pro:

[info] == put-many ==
[info] Running net.lag.kestrel.load.PutMany -n 1000000
Finished in 64921 msec (64.9 usec/put throughput).
Transactions: min=95.00; max=528107.00 524847.00 521780.00;
  median=3433.00; average=5551.77 usec
Transactions distribution: 5.00%=810.00 10.00%=1526.00 25.00%=2414.00
  50.00%=3433.00 75.00%=4851.00 90.00%=6933.00 95.00%=9145.00
  99.00%=59133.00 99.90%=208001.00 99.99%=505030.00
[info] == put-many ==

Many-clients

This test has one producer that trickles out one item at a time, and a pile of consumers fighting for each item. It usually takes exactly as long as the number of items times the delay, but is useful as a validation test to make sure kestrel works as advertised without blowing up.

$ sbt "many-clients --help"
usage: many-clients [options]
    spin up N clients and have them do timeout reads on a queue while a
    single producer trickles out.

options:
    -s MILLESCONDS
        sleep MILLISECONDS between puts (default: 100)
    -n ITEMS
        put ITEMS total items into the queue (default: 100)
    -c CLIENTS
        use CLIENTS consumers (default: 100)

A sample run on a 2010 MacBook Pro:

[info] == many-clients ==
[info] Running net.lag.kestrel.load.ManyClients
Finished in 11104 msec.
[info] == many-clients ==

Flood

This test starts up one producer and one consumer, and just floods items through kestrel as fast as it can.

$ sbt "flood --help"
usage: flood [options]
    spin up a producer and consumer and flood N items through kestrel

options:
    -n ITEMS
        put ITEMS items into the queue (default: 10000)
    -k KILOBYTES
        put KILOBYTES per queue item (default: 1)

A sample run on a 2010 MacBook Pro:

[info] == flood ==
[info] Running net.lag.kestrel.load.Flood -n 100000
flood: 100000 items of 1kB
Finished in 16834 msec (168.3 usec/put throughput).
Consumer spun 1 times in misses.

Packing

This test starts up one producer and one consumer, seeds the queue with a bunch of items to cause it to fall behind, then does cycles of flooding items through the queue, separated by pauses. It's meant to test kestrel's behavior with a queue that's fallen behind and stays behind indefinitely, to make sure the journal files are packed periodically without affecting performance too badly.

$ sbt "packing --help"
usage: packing [options]
    spin up a producer and consumer, write N items, then do read/write cycles
    with pauses

options:
    -q NAME
        use named queue (default: spam)
    -n ITEMS
        put ITEMS items into the queue (default: 25000)
    -k KILOBYTES
        put KILOBYTES per queue item (default: 1)
    -t SECONDS
        pause SECONDS between cycles (default: 1)
    -c CYCLES
        do read/writes CYCLES times (default: 100)
    -x
        use transactions when fetching

A sample run on a 2010 MacBook Pro:

[info] == packing ==
[info] Running net.lag.kestrel.load.JournalPacking -c 10 -q small
packing: 25000 items of 1kB with 1 second pauses
Wrote 25000 items starting at 0.
cycle: 1
Wrote 25000 items starting at 25000.
Read 25000 items in 5402 msec. Consumer spun 0 times in misses.
cycle: 2
Wrote 25000 items starting at 50000.
Read 25000 items in 5395 msec. Consumer spun 0 times in misses.
cycle: 3
Wrote 25000 items starting at 75000.
Read 25000 items in 4584 msec. Consumer spun 0 times in misses.
cycle: 4
Wrote 25000 items starting at 100000.
Read 25000 items in 4455 msec. Consumer spun 0 times in misses.
cycle: 5
Wrote 25000 items starting at 125000.
Read 25000 items in 4742 msec. Consumer spun 0 times in misses.
cycle: 6
Wrote 25000 items starting at 150000.
Read 25000 items in 4468 msec. Consumer spun 0 times in misses.
cycle: 7
Wrote 25000 items starting at 175000.
Read 25000 items in 5127 msec. Consumer spun 0 times in misses.
cycle: 8
Wrote 25000 items starting at 200000.
Read 25000 items in 4357 msec. Consumer spun 0 times in misses.
cycle: 9
Wrote 25000 items starting at 225000.
Read 25000 items in 4500 msec. Consumer spun 0 times in misses.
cycle: 10
Wrote 25000 items starting at 250000.
Read 25000 items in 4558 msec. Consumer spun 0 times in misses.
Read 25000 items in 3141 msec. Consumer spun 0 times in misses.
[info] == packing ==

You can see the journals being packed in the kestrel log:

INF [20110405-20:36:57.420] kestrel: Setting up queue small: maxItems=2147483647 maxSize=134217728.bytes maxItemSize=922
3372036854775807.bytes maxAge=None defaultJournalSize=16777216.bytes maxMemorySize=16777216.bytes maxJournalSize=1342177
28.bytes discardOldWhenFull=true keepJournal=true syncJournal=never expireToQueue=None maxExpireSweep=2147483647 fanoutO
nly=false
INF [20110405-20:36:57.421] kestrel: Replaying transaction journal for 'small'
INF [20110405-20:36:57.422] kestrel: No transaction journal for 'small'; starting with empty queue.
INF [20110405-20:36:57.422] kestrel: Finished transaction journal for 'small' (0 items, 0 bytes) xid=0
INF [20110405-20:36:59.779] kestrel: Rotating journal file for 'small' (qsize=16440320)
INF [20110405-20:36:59.852] kestrel: Dropping to read-behind for queue 'small' (16.0 MiB)
INF [20110405-20:37:02.032] kestrel: Rotating journal file for 'small' (qsize=29139968)
INF [20110405-20:37:04.583] kestrel: Rotating journal file for 'small' (qsize=35066880)
INF [20110405-20:37:05.005] kestrel: Read-behind on 'small' moving from file small.1302061022051 to small.1302061024673
INF [20110405-20:37:08.547] kestrel: Read-behind on 'small' moving from file small.1302061024673 to small
INF [20110405-20:37:09.553] kestrel: Rotating journal file for 'small' (qsize=27975680)
INF [20110405-20:37:12.412] kestrel: Read-behind on 'small' moving from file small.1302061029571 to small
INF [20110405-20:37:14.511] kestrel: Rotating journal file for 'small' (qsize=26700800)
INF [20110405-20:37:16.384] kestrel: Read-behind on 'small' moving from file small.1302061034588 to small
INF [20110405-20:37:17.122] kestrel: Rotating journal file for 'small' (qsize=29371392)
INF [20110405-20:37:20.164] kestrel: Read-behind on 'small' moving from file small.1302061037149 to small
INF [20110405-20:37:21.410] kestrel: Rotating journal file for 'small' (qsize=26664960)
INF [20110405-20:37:23.113] kestrel: Read-behind on 'small' moving from file small.1302061041427 to small
INF [20110405-20:37:25.302] kestrel: Rotating journal file for 'small' (qsize=26168320)
INF [20110405-20:37:27.118] kestrel: Read-behind on 'small' moving from file small.1302061045321 to small
INF [20110405-20:37:27.119] kestrel: Rewriting journal file from checkpoint for 'small' (qsize=27889664)
INF [20110405-20:37:27.129] kestrel: Packing journals for 'small': small.1302061019805, small.1302061022051, small.13020
61024673, small.1302061029571, small.1302061034588, small.1302061037149, small.1302061041427, small.1302061045321
INF [20110405-20:37:27.635] kestrel: Packing 'small' -- erasing old files.
INF [20110405-20:37:27.646] kestrel: Packing 'small' done: small.1302061045321, small
INF [20110405-20:37:28.115] kestrel: Rotating journal file for 'small' (qsize=28761088)
INF [20110405-20:37:31.108] kestrel: Read-behind on 'small' moving from file small.1302061048143 to small
INF [20110405-20:37:32.202] kestrel: Rotating journal file for 'small' (qsize=27242496)
INF [20110405-20:37:34.048] kestrel: Read-behind on 'small' moving from file small.1302061052221 to small
INF [20110405-20:37:36.255] kestrel: Rotating journal file for 'small' (qsize=25759744)
INF [20110405-20:37:38.433] kestrel: Read-behind on 'small' moving from file small.1302061056360 to small
INF [20110405-20:37:39.550] kestrel: Rotating journal file for 'small' (qsize=27325440)
INF [20110405-20:37:42.266] kestrel: Read-behind on 'small' moving from file small.1302061059646 to small
INF [20110405-20:37:43.464] kestrel: Rotating journal file for 'small' (qsize=26256384)
INF [20110405-20:37:45.110] kestrel: Read-behind on 'small' moving from file small.1302061063469 to small
INF [20110405-20:37:46.110] kestrel: Rotating journal file for 'small' (qsize=27487232)
INF [20110405-20:37:48.928] kestrel: Read-behind on 'small' moving from file small.1302061066128 to small
INF [20110405-20:37:49.875] kestrel: Rotating journal file for 'small' (qsize=28101632)
INF [20110405-20:37:51.801] kestrel: Read-behind on 'small' moving from file small.1302061069893 to small
INF [20110405-20:37:51.801] kestrel: Rewriting journal file from checkpoint for 'small' (qsize=26379264)
INF [20110405-20:37:51.804] kestrel: Packing journals for 'small': small.1302061045321, small.1302061048143, small.13020
61052221, small.1302061056360, small.1302061059646, small.1302061063469, small.1302061066128, small.1302061069893
INF [20110405-20:37:52.237] kestrel: Packing 'small' -- erasing old files.
INF [20110405-20:37:52.246] kestrel: Packing 'small' done: small.1302061069893, small
INF [20110405-20:37:54.012] kestrel: Rotating journal file for 'small' (qsize=26510336)
INF [20110405-20:37:55.808] kestrel: Read-behind on 'small' moving from file small.1302061074039 to small
INF [20110405-20:37:56.594] kestrel: Rotating journal file for 'small' (qsize=29006848)
INF [20110405-20:37:59.363] kestrel: Read-behind on 'small' moving from file small.1302061076614 to small
INF [20110405-20:37:59.731] kestrel: Coming out of read-behind for queue 'small'