NAME
stompserver - Stomp protocol messaging server
SYNOPSIS
stompserver [options]
DESCRIPTION
Stomp messaging server with file/dbm/memory/activerecord based FIFO
queues, queue monitoring, and basic authentication.
OPTIONS
-C, --config=CONFIGFILE
Configuration File (default: stompserver.conf)
-p, --port=PORT
Change the port (default: 61613)
-b, --host=ADDR
Change the host (default: localhost)
-q, --queuetype=QUEUETYPE
Queue type (memory|dbm|activerecord|file) (default: memory)
-w, --working_dir=DIR
Change the working directory (default: current directory)
-s, --storage=DIR
Change the storage directory (default: .stompserver, relative to
working_dir)
-d, --debug
Turn on debug messages
-a, --auth
Require client authorization
-c, --checkpoint=SECONDS
Time between checkpointing the queues in seconds (default: 0)
-h, --help
Show this message
QUEUES
Stompserver handles basic message queue processing using memory, file,
or dbm based queues. Messages are sent and consumed in FIFO order
(unless a client error happens, this should be corrected in the
future). Topics are memory-only storage. You can select activerecord,
file or dbm storage and the queues will use that, but topics will only
be stored in memory.
memory queues are of course the fastest ones but shouldn’t be used if
you want to ensure all messages are delivered.
dbm queues will use berkeleydb if available, otherwise dbm or gdbm
depending on the platform. sdbm does not work well with marshalled
data. Note that these queues have not been tested in this release.
For the file based storage, each frame is stored in a single file. The
first 8 bytes contains the header length, the next 8 bytes contains the
body length, then the headers are stored as a marshalled object
followed by the body stored as a string. This storage is currently
inefficient because queues are stored separately from messages, which
forces a double write for data safety reasons on each message stored.
The activerecord based storage expects to find a database.yml file in
the configuration directory. It should be the most robust backend, but
the slowest one. The database must have an ar_messages table which can
be created with the following code (you are responsible to do so):
ActiveRecord::Schema.define do
create_table ’ar_messages’ do |t|
t.column ’stomp_id’, :string, :null => false
t.column ’frame’, :text, :null => false
end
end
You can read the frames with this model:
class ArMessage < ActiveRecord::Base
serialize :frame
end
The ar_message implementation will certainly change in the future.
This is meant to be easily readable by a Rails application (which could
handle the ar_messages table creation with a migration).
ACCESS CONTROL
Basic client authorization is also supported. If the -a flag is passed
to stompserver on startup, and a .passwd file exists in the run
directory, then clients will be required to provide a valid login and
passcode. See passwd.example for the password file format.
MONITORING
Queues can be monitored via the monitor queue (this will probably not
be supported this way in the future to avoid polluting the queue
namespace). If you subscribe to /queue/monitor, you will receive a
status message every 5 seconds that displays each queue, it’s size,
frames enqueued, and frames dequeued. Stats are sent in the same format
of stomp headers, so they are easy to parse. Following is an example of
a status message containing stats for 2 queues:
Queue: /queue/client2 size: 0 dequeued: 400 enqueued: 400
Queue: /queue/test size: 50 dequeued: 250 enqueued: 300
AUTHOR
stompserver was written by Patrick Hurley <phurley@gmail.com> and
Lionel Bouton.
This manual page was compiled from the included documentation by Bryan
McLellan <btm@loftninjas.org> for the Debian project (and may be used
by others). The existing documentation is distributed under the MIT
license.