| Manual section: | 7 |
|---|---|
| Date: | July 2017 |
| Version: | 2.17.07a2 |
| Manual group: | Metpx-Sarracenia Suite |
- sr_component [foreground|start|stop|restart|status|cleanup|declare|setup] <config>
- <config_dir>/ [ default.conf ]
- <config_dir>/ [ sarra | subscribe | report | sender | watch ] / <config.conf>
- <config_dir>/ scripts / <script.py>
Metpx Sarracenia components are the programs that can be invoked from the command line: examples are: sr_subscribe, sr_sarra, sr_sender, sr_report, When any component is invoked, a configuration file and an operation are specified. The operation is one of:
- foreground: run a single instance in the foreground logging to stderr
- restart: stop and then start the configuration.
- start: start the configuration running
- status: check if the configuration is running.
- stop: stop the configuration from running
The remaining operations manage the resources (exchanges,queues) used by the component on the rabbitmq server.
- cleanup: deletes the component's resources on the server
- declare: creates the component's resources on the server
- setup: like declare, additionnaly does queue bindings
For example: sr_subscribe foreground dd runs the sr_subcribe component with the dd configuration as a single foreground instance.
The foreground action is used when building a configuration or for debugging. The foreground instance will run regardless of other instances which are currently running. Should instances be running, it shares the same message queue with them. A user stop the foreground instance by simply using <ctrl-c> on linux or use other means to kill the process.
help has a component print a list of valid options.
FIXME: Cannot find a component where help works. Remove? FIXME: OK: sr_post -h works. but not sr_post 'help' (interpreted as a file, naturally.)
-V has a component print out a version identifier and exit.
Metpx Sarracenia is configured using a tree of text files using a common syntax. The location of config dir is platform dependent (see python appdirs):
- linux: ~/.config/sarra - Windows: %AppDir%/science.gc.ca/sarra, this might be: C:\Users\peter\AppData\Local\science.gc.ca\sarra
The top of the tree contains a file 'default.conf' which contains settings that are read as defaults for any component on start up. Individual configuration files can be placed anywhere and invoked with the complete path. When components are invoked, the provided file is interpreted as a file path (with a .conf suffix assumed.) If it is not found as file path, then the component will look in the component's config directory ( config_dir / component ) for a matching .conf file.
If it is still not found, it will look for it in the site config dir (linux: /usr/share/default/sarra/component).
Finally, if the user has set option remote_config to True and if he has configured web sites where configurations can be found (option remote_config_url), The program will try to download the named file from each site until it finds one. If successful, the file is downloaded to config_dir/Downloads and interpreted by the program from there. There is a similar process for all plugins that can be interpreted and executed within sarracenia components. Components will first look in the plugins directory in the users config tree, then in the site directory, then in the sarracenia package itself, and finally it will look remotely.
FIXME provide some sample file locations FIXME network search path...
Options are placed in configuration files, one per line, in the form:
option <value>
For example:
**debug true**
sets the debug option to enable more verbose logging. To provide non-functional description of configuration, or comments, use lines that begin with a #.
All options are case sensitive. Debug is not the same as debug or DEBUG. Those are three different options (two of which do not exist and will have no effect, but should generate an ´unknown option warning´.)
Options and command line arguments are equivalent. Every command line argument has a corresponding long version starting with '--'. For example -u has the long form --url. One can also specify this option in a configuration file. To do so, use the long form without the '--', and put its value separated by a space. The following are all equivalent:
- url <url>
- -u <url>
- --url <url>
Settings in an individual .conf file are read in after the default.conf file, and so can override defaults. Options specified on the command line override configuration files.
Settings are interpreted in order. Each file is read from top to bottom. for example:
sequence #1:
reject .*\.gif accept .*
sequence #2:
accept .* reject .*\.gif
FIXME: does this match only files ending in 'gif' or should we add a $ to it? will it match something like .gif2 ? is there an assumed .* at the end?
In sequence #1, all files ending in 'gif' are rejected. In sequence #2, the accept .* (which accepts everything) is encountered before the reject statement, so the reject has no effect.
Several options that need to be reused in different config file can be grouped in a file. In each config where the options subset should appear, the user would then use :
- --include <includeConfigPath>
The includeConfigPath would normally reside under the same config dir of its master configs. There is no restriction, any option can be placed in a config file included. The user must be aware that, for most options, several declarations means overwriting their values.
Usernames and passwords or keys used to access servers are examples of credentials. In order to reduce the sensitivity of most configuration files, the credentials are stored in a single file apart from all other settings. The credentials.conf file is the only mandatory configuration file for all users.
For all sarracenia programs, the confidential parts of credentials are stored only in ~/.config/sarra/credentials.conf. This includes the destination and the broker passwords and settings needed by components. The format is one entry per line. Examples:
In other configuration files or on the command line, the url simply lacks the password or key specification. The url given in the other files is looked up in credentials.conf.
Most Metpx Sarracenia components loop on reception and consumption of sarracenia AMQP messages. Usually, the messages of interest are sr_post messages, announcing the availability of a file by publishing it´s URL ( or a part of a file ), but there are also sr_report(7) messages which can be processed using the same tools. AMQP messages are published to an exchange on a broker (AMQP server.) The exchange delivers messages to queues. To receive messages, one must provide the credentials to connect to the broker (AMQP message pump). Once connected, a consumer needs to create a queue to hold pending messages. The consumer must then bind the queue to one or more exchanges so that they put messages in its queue.
Once the bindings are set, the program can receive messages. When a message is received, further filtering is possible using regular expression onto the AMQP messages. After a message passes this selection process, and other internal validation, the component can run an on_message plugin script to perform additional message processing. If this plugin returns False, the message is discarded. If True, processing continues.
The following sections explains all the options to set this "consuming" part of sarracenia programs.
broker amqp{s}://<user>:<password>@<brokerhost>[:port]/<vhost>
An AMQP URI is used to configure a connection to a message pump (aka AMQP broker.) Some sarracenia components set a reasonable default for that option. You provide the normal user,host,port of connections. In most configuration files, the password is missing. The password is normally only included in the credentials.conf file.
Sarracenia work has not used vhosts, so vhost should almost always be /.
for more info on the AMQP URI format: ( https://www.rabbitmq.com/uri-spec.html )
either in the default.conf or each specific configuration file. The broker option tell each component which broker to contact.
broker amqp{s}://<user>:<pw>@<brokerhost>[:port]/<vhost>
Once connected to an AMQP broker, the user needs to bind a queue to exchanges and topics to determine the messages of interest.
Usually components guess reasonable defaults for all these values and users do not need to set them. For less usual cases, the user may need to override the defaults. The queue is where the notifications are held on the server for each subscriber.
By default, components create a queue name that should be unique. The default queue_name components create follows : q_<brokerUser>.<programName>.<configName> . Users can override the defaul provided that it starts with q_<brokerUser>. Some variables can also be used within the queue_name like ${BROKER_USER},${PROGRAM},${CONFIG},${HOSTNAME}
The durable option, if set to True, means writes the queue on disk if the broker is restarted.
The expire option is expressed in minutes... it sets how long should live a queue without connections The durable option set to True, means writes the queue on disk if the broker is restarted.
The message-ttl option set the time in minutes a message can live in the queue. Past that time, the message is taken out of the queue by the broker.
The prefetch option sets the number of messages to fetch at one time. When multiple instances are running and prefetch is 4, each instance will obtain upto four messages at a time. To minimize the number of messages lost if an instance dies and have optimal load sharing, the prefetch should be set as low as possible. However, over long haul links, it is necessary to raise this number, to hide round-trip latency, so a setting of 10 or more may be needed.
When --reset is set, and a component is (re)started, its queue is deleted (if it already exists) and recreated according to the component's queue options. This is when a broker option is modified, as the broker will refuse access to a queue declared with options that differ from what was set at creation. It can also be used to discard a queue quickly when a receiver has been shut down for a long period.
The AMQP protocol defines other queue options which are not exposed via sarracenia, because sarracenia itself picks appropriate values.
Users almost always need to set these options. Once a queue exists on the broker, it must be bound to an exchange. Bindings define which messages (URL notifications) the program receives. The root of the topic tree is fixed to indicate the protocol version and type of the message (but developers can override it with the topic_prefix option.)
So the binding options are:
- exchange <name> (default: xpublic)
- topic_prefix <amqp pattern> (default: varies by component)
- subtopic <amqp pattern> (subtopic need to be set)
Usually, the user specifies one exchange, and several subtopic options. Subtopic is what is normally used to indicate messages of interest. To use the subtopic to filter the products, match the subtopic string with the relative path of the product.
For example, consuming from DD, to give a correct value to subtopic, one can browse the our website http://dd.weather.gc.ca and write down all directories of interest. For each directory tree of interest, write a subtopic option as follow:
subtopic directory1.*.subdirectory3.*.subdirectory5.#
where:
* replaces a directory name
# stands for the remaining possibilities
One has the choice of filtering using subtopic with only AMQP's limited wildcarding and header length limited to 255 encoded bytes, or the more powerful regular expression based accept/reject mechanisms described below, which are not length limited. The difference being that the AMQP filtering is applied by the broker itself, saving the notices from being delivered to the client at all. The accept/reject patterns apply to messages sent by the broker to the subscriber. In other words, accept/reject are client side filters, whereas subtopic is server side filtering.
It is best practice to use server side filtering to reduce the number of announcements sent to the client to a small superset of what is relevant, and perform only a fine-tuning with the client side mechanisms, saving bandwidth and processing for all.
topic_prefix is primarily of interest during protocol version transitions, where one wishes to specify a non-default protocol version of messages to subscribe to.
We have selected our messages through exchange, subtopic and perhaps patterned subtopic with AMQP's limited wildcarding. The broker puts the corresponding messages in our queue. The component downloads the these messages.
Sarracenia clients implement a the more powerful client side filtering using regular expression based mechanisms.
The accept and reject options use regular expressions (regexp). The regexp is applied to the the message's URL for a match.
If the message's URL of a file matches a reject pattern, the message is acknowledged as consumed to the broker and skipped.
One that matches an accept pattern is processed by the component.
In many configurations, accept and reject options are mixed with the directory option. They then relate accepted messages to the directory value they are specified under.
After all accept / reject options are processed, normally the message acknowledged as consumed and skipped. To override that default, set accept_unmatch to True. However, if no accept / reject are specified, the program assumes it should accept all messages and sets accept_unmatch to True.
The accept/reject are interpreted in order. Each option is processed orderly from top to bottom. for example:
sequence #1:
reject .*\.gif accept .*
sequence #2:
accept .* reject .*\.gif
In sequence #1, all files ending in 'gif' are rejected. In sequence #2, the accept .* (which accepts everything) is encountered before the reject statement, so the reject has no effect.
It is best practice to use server side filtering to reduce the number of announcements sent to the component to a small superset of what is relevant, and perform only a fine-tuning with the client side mechanisms, saving bandwidth and processing for all.
Once a message has gone through the filtering above, the user can run plugins on the message and perform arbitrary processing (in Python 3.) For example: to do statistics, rename a product, changing its destination... example usage:
The on_message plugin scripts are called as the last step in consuming messages. All plugin scripts return a boolean. If False is returned, the component acknowledges the message to the broker and does not process it. If no on_message plugin is set, or if all plugins provided returned True, the message is processed by the component.
Sources of data need to indicate the clusters to which they would like data to be delivered. Data Pumps need to identify themselves, and their neighbors in order to pass data to them.
Components which inject data into a network (sr_post, sr_poll, sr_watch) need to set 'to' addresses for all data injected. Components which transfer data between pumps, such as sr_sarra and sr_sender, interpret cluster, cluster_aliases, and gateway_for, such that products which are not meant for the destination cluster are not transferred.
The network will not process a message that
1- has no source (message.headers['source']) 2- has no origin (message.headers['from_cluster']) 3- has no destination (message.headers['to_clusters']) (**to** option on post/watch/poll) 4- the to_clusters destination list has no match with this pump's **cluster,cluster_aliases,gateway_for** options
FIXME: all of the above, I'm a bit confused about, explanation seems complicated need to rephrase...
These options set what files will be downloaded, where they will be placed, and under which name.
The attempts option sets how many times to try to download a product.
The batch option is used to indicate how many files should be transferred over a connection, before it is torn down, and re-established. On very low volume transfers, where timeouts can occur between transfers, this should be lowered to 1. For most usual situations the default is fine. for higher volume cases, one could raise it to reduce transfer overhead. It is only used for file transfer protocols, not HTTP ones at the moment.
The heartbeat option sets how often to execute periodic processing as determined by the list of on_heartbeat plugins. By default, it prints a log message every heartbeat.
The inflight option sets how to ignore files when they are being transferred or (in mid-flight betweeen two systems.) This assures that consuming programs reading the directory ignore them until they are ready. The value can be a file name suffix, which is appended to create a temporary name during the transfer. If inflight is set to ., then it is prefix, to conform with the standard for "hidden" files on unix/linux. In either case, when the transfer is complete, the file is renamed to it's permanent name to allow further processing.
If inflight is set to an integer value, then it specifies the minimum number of seconds a file must have remained unchanged before it is accepted for processing (used by sr_watch)
Directory sets where to put the files on your server. Combined with accept / reject options, the user can select the files of interest and their directories of residence.
ex. directory /mylocaldirectory/myradars
accept .*RADAR.*
directory /mylocaldirectory/mygribs
reject .*Reg.*
accept .*GRIB.*
The mirror option can be used to mirror the dd.weather.gc.ca tree of the files. If set to True the directory given by the directory option will be the basename of a tree. Accepted files under that directory will be placed under the subdirectory tree leaf where it resides under dd.weather.gc.ca. For example retrieving the following url, with options:
http://dd.weather.gc.ca/radar/PRECIP/GIF/WGJ/201312141900_WGJ_PRECIP_SNOW.gif mirror True directory /mylocaldirectory accept .*RADAR.*
would result in the creation of the directories and the file /mylocaldirectory/radar/PRECIP/GIF/WGJ/201312141900_WGJ_PRECIP_SNOW.gif
Use the option strip set to N (an integer) to trim the beginnning of the directory tree. For example:
http://dd.weather.gc.ca/radar/PRECIP/GIF/WGJ/201312141900_WGJ_PRECIP_SNOW.gif mirror True strip 3 directory /mylocaldirectory accept .*RADAR.*
would result in the creation of the directories and the file /mylocaldirectory/WGJ/201312141900_WGJ_PRECIP_SNOW.gif, stripping out radar, PRECIP, and GIF from the path.
The flatten option is use to set a separator character. This character replaces the '/' in the url to create a "flattened" filename from its dd.weather.gc.ca path. For example, retrieving the following url with options:
http://dd.weather.gc.ca/model_gem_global/25km/grib2/lat_lon/12/015/CMC_glb_TMP_TGL_2_latlon.24x.24_2013121612_P015.grib2 flatten - directory /mylocaldirectory accept .*model_gem_global.*
results in the creating
/mylocaldirectory/model_gem_global-25km-grib2-lat_lon-12-015-CMC_glb_TMP_TGL_2_latlon.24x.24_2013121612_P015.grib2
The overwrite option, when set, forces overwriting of an existing file even if it has the same checksum as the newly advertised version.
The filename option can be used to completely overwrite the delivery file name. The regexp pattern in an accept clause can be used to set numbered fields for use in subsequently building a destination path by placing a portion of the regex in parentheses. One can use these numbered fields to build the directory name. Example, in the configuration file:
directory /this/${0}/pattern/${1}/directory
accept .*(2016....).*(RAW.*GRIB).*
if the message receiving a notice like:
20150813161959.854 http://this.pump.com/ relative/path/to/20160123_product_RAW_MERGER_GRIB_from_CMC
would result in the file being placed as follows:
/this/20160123/pattern/RAW_MERGER_GRIB/directory/20160123_product_RAW_MERGER_GRIB_from_CMC
FIXME: Is it correct for this to be different for sr_subscribe? why is default not False everywhere?
kbytes_ps is greater than 0, the process attempts to respect this delivery speed in kilobytes per second... ftp,ftps,or sftp)
Permission bits on the destination files written are controlled by the mode directives. preserve_modes will apply the mode permissions posted by the source of the file. If no source mode is available, the default_mode will be applied to files, and the default_dir_mode will be applied to directories.
destfn_script <script> (default:None)
This destination file name script option defines a plugin to be run when ready to deliver a product. The script receives the component class instance (same as an on_* plugin), with access to all its variables. Any modification to parent.remote_file, for example, will set a new destination filename
filename <keyword> (default:WHATFN)
The support of this option give all sorts of possibilities for setting the remote filename. Some keywords are based on the fact that metpx-sundew filenames are five (to six) fields strings separated by for colons. The possible keywords are:
accept <regexp pattern> [<keyword>]
keywords can also be added as a second argument to the accept option. The keyword is any one of the filename option. A message that matched against the accept regexp pattern, will have its remote_file modified as per this keyword option. This keyword has priority over the preceeding filename.
Examples:
filename NONE
directory /this/first/target/directory
accept .*file.*type1.*
directory /this/target/directory
accept .*file.*type2.*
accept .*file.*type3.* DESTFN=file_of_type3
directory /this/${0}/pattern/${1}/directory
accept .*(2016....).*(RAW.*GRIB).*
A selected message by the first accept would be delivered unchanged to the first directory. A selected message by the second accept would be delivered unchanged to the second directory. A selected message by the third accept would be renamed "file_of_type3" in the second directory. A selected message by the forth accept would be delivered unchanged to a directory
named /this/20160123/pattern/RAW_MERGER_GRIB/directory if the message would have a notice like :
20150813161959.854 http://this.pump.com/ relative/path/to/20160123_product_RAW_MERGER_GRIB_from_CMC
In MetPX Sundew, there is a much more strict file naming standard, specialised for use with World Meteorological Organization (WMO) data. Note that the file naming convention predates, and bears no relation to the WMO file naming convention currently approved, but is strictly an internal format. The files are separated into six fields by colon characters. The first field, DESTFN, gives the WMO (386 style) Abbreviated Header Line (AHL) with underscores replacing blanks: TTAAii CCCC YYGGGg BBB ... (see WMO manuals for details) followed by numbers to render the product unique (as in practice, though not in theory, there are a large number of products which have the same identifiers.) The meanings of the fifth field is a priority, and the last field is a date/time stamp. A sample file name:
SACN43_CWAO_012000_AAA_41613:ncp1:CWAO:SA:3.A.I.E:3:20050201200339
If a file is sent to sarracenia and it is named according to the sundew conventions, then the following substition fields are available:
${T1} replace by bulletin's T1
${T2} replace by bulletin's T2
${A1} replace by bulletin's A1
${A2} replace by bulletin's A2
${ii} replace by bulletin's ii
${CCCC} replace by bulletin's CCCC
${YY} replace by bulletin's YY (obs. day)
${GG} replace by bulletin's GG (obs. hour)
${Gg} replace by bulletin's Gg (obs. minute)
${BBB} replace by bulletin's bbb
${RYYYY} replace by reception year
${RMM} replace by reception month
${RDD} replace by reception day
${RHH} replace by reception hour
${RMN} replace by reception minutes
${RSS} replace by reception second
The 'R' fields from from the sixth field, and the others come from the first one.
Components write to log files, which by default are found in ~/.cache/sarra/var/log/<component>_<config>_<instance>.log. at the end of the day, These logs are rotated automatically by the components, and the old log gets a date suffix. The directory in which the logs are stored can be overridden by the log option, and the number of days' logs to keep is set by the 'logdays' parameter. Log files older than logdays days are deleted.
debug setting option debug is identical to use loglevel debug
log the directory to store log files in. Default value: ~/.cache/sarra/var/log (on Linux)
logdays the number of days' log files to keep online, assuming a daily rotation.
possible values are : critical, error, info, warning, debug.
chmod_log the permission bits to set on log files (default 0600 )
Note: for sr-post only, option log should be a logfile
FIXME: I don't understand the point of logging a post... it seems like it should always be 'foreground' and that it would just write to stderr... it is a one-time thing... confused. what would it log?
FIXME: We need a verbosity setting. should probably be documented here. on INFO, the logs are way over the top verbose. Probably need to trim that down. log_level?
Sometimes one instance of a component and configuration is not enough to process & send all available notifications.
instances <integer> (default:1)
The instance option allows launching serveral instances of a component and configuration. When running sr_sender for example, a number of runtime files that are created. In the ~/.cache/sarra/sender/configName directory:
A .sr_sender_configname.state is created, containing the number instances. A .sr_sender_configname_$instance.pid is created, containing the PID of $instance process.
In directory ~/.cache/sarra/var/log:
A .sr_sender_configname_$instance.log is created as a log of $instance process.
The logs can be written in another directory than the default one with option :
log <directory logpath> (default:~/.cache/sarra/var/log)
FIXME: indicate windows location also... dot files on windows?
While the brokers keep the queues available for some time, Queues take resources on brokers, and are cleaned up from time to time. A queue which is not accessed and has too many (implementation defined) files queued will be destroyed. Processes which die should be restarted within a reasonable period of time to avoid loss of notifications. A queue which is not accessed for a long (implementation dependent) period will be destroyed.
These options are only used when more than one broker needs to be configured for a component, ( such as sr_sarra(8), sr_sender(1), sr_shovel(1), sr_winnow(1).) These options specify the broker to which messages are output, or "posted." By default, components publishes the selected consumed message with its exchange onto the current cluster, with the feeder account.
The user can overwrite the defaults with options :
The post_broker option sets the credential informations to connect to the output RabbitMQ server. The default is the value of the feeder option in default.conf.
The post_exchange option sets a new exchange for the selected messages. The default is to publish under the exchange it was consumed. Before a message is published, a user can set to trigger a script. The option on_post would be used to do such a setup. If the script returns True, the message is published... and False it wont.
The post_exchange_split option is EXPERIMENTAL. It appends a two digit suffix resulting from hashing the last character of the checksum to the post_exchange name, in order to divide the output amongst a number of exchanges. This is currently used in high traffic pumps to allow multiple instances of sr_winnow, which cannot be instanced in the normal way. example:
post_exchange_split 5 post_exchange xwinnow
will result in posting messages to five exchanges named: xwinnow00, xwinnow01, xwinnow02, xwinnow03 and xwinnow04, where each exchange will receive only one fifth of the total flow.
For each download, an amqp report message is sent back to the broker. Should you want to turned them off the option is :
When a report is generated, it is sent to the configured report_exchange. Administrive components post directly to xreport, whereas user components post to their own exchanges. The report daemons copy the messages to xreport after validation.
The feeder option specifies the account used by default system transfers for components such as sr_shovel, sr_sarra and sr_sender (when posting).
When set, the admin option will cause sr start to start up the sr_audit daemon.
FIXME:: feeder is perhaps an archeological artifact. perhaps should disappear and just use broker for this, when run as the admin user. then the trigger to run all admin daemons would be the presence of the admin user in the configuration.
Most users are defined using the declare option.
Role:
subscriber
A subscriber is user that can only subscribe to data and return report messages. Subscribers are not permitted to inject data. Each subscriber has an xs_<user> named exchange on the pump, where if a user is named Acme, the corresponding exchange will be xs_Acme. This exchange is where an sr_subscribe process will send it's report messages.
By convention/default, the anonymous user is created on all pumps to permit subscription without a specific account.
source
A user permitted to subscribe or originate data. A source does not necessarily represent one person or type of data, but rather an organization responsible for the data produced. So if an organization gathers and makes available ten kinds of data with a single contact email or phone number for questions about the data and it's availability, then all of those collection activities might use a single 'source' account.
Each source gets a xs_<user> exchange for injection of data posts, and, similar to a subscriber to send report messages about processing and receipt of data. source may also have an xl_<user> exchange where, as per report routing configurations, report messages of consumers will be sent.
Metpx Sarracenia provides minimum functionality to deal with the most common cases, but provides flexibility to override those common cases with user plugins scripts, written in python. MetPX comes with a variety of scripts which can act as examples.
Users can place their own scripts in the script sub-directory of their config directory tree.
A user script should be placed in the ~/.config/sarra/plugins directory.
There are two varieties of scripts: do_* and on_*. Do_* scripts are used to implement functions, replacing built-in functionality, for example, to implement additional transfer protocols.
On_* plugins are used more often. They allow actions to be inserted to augment the default processing for various specialized use cases. The scripts are invoked by having a given configuration file specify an on_<event> option. The event can be one of:
The simplest example of a plugin: A do_nothing.py script for on_file:
class Transformer(object):
def __init__(self):
pass
def perform(self,parent):
logger = parent.logger
logger.info("I have no effect but adding this log line")
return True
transformer = Transformer()
self.on_file = transformer.perform
The only arguments the script receives it parent, which is an instance of the sr_subscribe class Should one of these scripts return False, the processing of the message/file will stop there and another message will be consumed from the broker. For other events, the last line of the script must be modified to correspond. Multiple scripts can be attached to the same event, in which case they are called in the order they appear in the configuration.
One can specify on_message None to reset the list to no plugins (removes msg_log, so it suppresses logging of message receipt.)
More examples are available in the Programming Guide.
Here is a short complete example configuration file:
broker amqp://dd.weather.gc.ca/ subtopic model_gem_global.25km.grib2.# accept .*
This above file will connect to the dd.weather.gc.ca broker, connecting as anonymous with password anonymous (defaults) to obtain announcements about files in the http://dd.weather.gc.ca/model_gem_global/25km/grib2 directory. All files which arrive in that directory or below it will be downloaded into the current directory (or just printed to standard output if -n option was specified.)
A variety of example configuration files are available here:
http://sourceforge.net/p/metpx/git/ci/master/tree/sarracenia/samples/config/
sr_report(7) - the format of report messages.
sr_report(1) - process report messages.
sr_post(1) - post announcemensts of specific files.
sr_post(7) - The format of announcement messages.
sr_sarra(1) - Subscribe, Acquire, and ReAdvertise tool.
sr_watch(1) - the directory watching daemon.
http://metpx.sf.net/ - sr_subscribe is a component of MetPX-Sarracenia, the AMQP based data pump.