 
.. This work is licensed under a Creative Commons Attribution 4.0 International License. 
.. SPDX-License-Identifier: CC-BY-4.0 
.. CAUTION: this document is generated from source in doc/src/rtd. 
.. To make changes edit the source and recompile the document. 
.. Do NOT make changes directly to .rst or .md files. 
 
 
RMR Overview 
============================================================================================ 
 
 
RMr Library 
============================================================================================ 
 
 
NAME 
-------------------------------------------------------------------------------------------- 
 
RMr -- Ric Message Router Library 
 
DESCRIPTION 
-------------------------------------------------------------------------------------------- 
 
RMr is a library which provides a user application with the 
ability to send and receive messages to/from other RMr based 
applications without having to understand the underlying 
messaging transport environment (e.g. Nanomsg) and without 
needing to know which other endpoint applications are 
currently available and accepting messages. To do this, RMr 
depends on a routing table generated by an external source. 
This table is used to determine the destination endpoint of 
each message sent by mapping the message type T (supplied by 
the user application) to an endpoint entry. Once determined, 
the message is sent directly to the endpoint. The user 
application is unaware of which endpoint actually receives 
the message, and in some cases whether that message was sent 
to multiple applications. 
 
RMr functions do provide for the ability to respond to the 
specific source instance of a message allowing for either a 
request response, or call response relationship when needed. 
 
The Route Table 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 
 
The library must be given a route table which maps message 
numbers to endpoint groups such that each time a message of 
type T is sent, the message is delivered to one member of 
each group associated with T. For example, message type 2 
might route to two different groups where group A consists of 
worker1 and worker2, while group B consists only of logger1. 
 
It is the responsibility of the route table generator to know 
which endpoints belong to which groups, and which groups 
accept which message types. Once understood, the route table 
generator publishes a table that is ingested by RMr and used 
for mapping messages to end points. 
 
The following is a simple route table which causes message 
types 0 through 9 to be routed to specific applications: 
 
:: 
  
 newrt|start
    mse|0|-1| %meid
    mse|1|-1|app10:4560,app11:4560
    mse|2|-1|app12:4560
    mse|3|-1|app14:4560
    mse|4|-1|app18:4560
    mse|5|-1|app01:4560
    mse|6|-1|app02:4560
    mse|7|-1|app03:4560
    mse|8|-1|app04:4560
    mse|9|-1|app05:4560
 newrt|end
 
 
 
The special endpoint "%meid" indicates that the message type 
(0 in this case) is to be routed to the endpoint which has 
been listed as the "owner" for the meid appearing in the 
message. MEID ownership is communicated to RMR using the same 
Route Table Manager interface and by supplying a "table" such 
as the one below: 
 
:: 
  
 meid_map | start
    mme_ar | control1 | meid000 meid001 meid002 meid003 meid004 meid005
    mme_ar | control2 | meid100 meid101 meid102 meid103
 meid_map | end | 2
 
 
This table indicates that the application (endpoint) 
*control1* "owns" 6 MEIDs and *control2* owns 4. When message 
type 0 is sent, the MEID in the message will be used to 
select the endpoint via this table. 
 
The MEID table will update the existing owner relationships, 
and add new ones; it is necessary to send only the changes 
with the add/replace (mme_ar) entries in the table. When 
necessary, MEIDs can be deleted by adding an mme_del record 
to the table. The following example illustrates how this 
might look: 
 
:: 
  
 meid_map | start
    mme_ar | control1 | meid000 meid001 meid002 meid003 meid004 meid005
    mme_ar | control2 | meid100 meid101 meid102 meid103
    mme_del| meid200 meid401
 meid_map | end | 3
 
 
 
Route Table Syntax 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 
 
The following illustrates the syntax for both the route 
table. 
 
 
:: 
  
 newrt | start
 mse | <message-type>[,<sender-endpoint>] | <sub-id> <roud-robin-grp>[;<round-robin-grp>]...
 newrt | end
 
 
 
A round robin group is one or more endpoints from which one 
will be selected to receive the message. When multiple 
endpoints are given in a group, they must be separated with a 
comma. An endpoint is the IP address and port (e.g. 
192.158.4.30:8219) or DNS name and port of the application 
that should receive the message type. If multiple round-robin 
groups are given, they must be separated by a semicolon, and 
 
MEID Map Syntax 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 
 
The MEID map is similar to the route table. Entries are used 
to add or replace the ownership of one or more MEIDs (mme_ar) 
or to delete one or more MEIDs (mme_del). The following is 
the syntax for the MEID map. 
 
 
:: 
  
 meid_map | start
 mme_ar | <owner-endpoint> | <meid> [<meid>...]
 mme_del | <meid> [<meid>...]
 meid_map | end | <count> [| <md5sum>
 
 
 
The <count> on the end record indicates the number of mme_ar 
and mme_del records which were sent; if the count does not 
match the whole map is refused and dropped. The 
<owner-endpoint> is the endpoint which should receive the 
message when a message is routed based on the MEID it 
contains. A MEID may be "owned" by only one endpoint, and if 
supplied multiple times, the last observed relationship is 
used. Each of the lists of MEIDs are blank separated. 
 
The optional <md5sum> on the *end* record should be the 
computed MD5 hash for all records which appear between the 
start and and records. This allows for a tighter verification 
that all data was received exactly as the route manager 
transmitted them. 
 
Environment 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 
 
To enable configuration of the library behaviour outside of 
direct user application control, RMr supports a number of 
environment variables which provide information to the 
library. The following is a list of the various environment 
variables, what they control and the defaults which RMr uses 
if undefined. 
 
 
 
RMR_ASYNC_CONN 
   
  Allows the asynch connection mode to be turned off (by 
  setting the value to 0. When set to 1, or missing from the 
  environment, RMR will invoke the connection interface in 
  the transport mechanism using the non-blocking (asynch) 
  mode. This will likely result in many "soft failures" 
  (retry) until the connection is established, but allows 
  the application to continue unimpeeded should the 
  connection be slow to set up. 
 
RMR_BIND_IF 
   
  This provides the interface that RMr will bind listen 
  ports to allowing for a single interface to be used rather 
  than listening across all interfaces. This should be the 
  IP address assigned to the interface that RMr should 
  listen on, and if not defined RMr will listen on all 
  interfaces. 
 
RMR_RTG_SVC 
   
  RMr opens a TCP listen socket using the port defined by 
  this environment variable and expects that the route table 
  generator process will connect to this port. If not 
  supplied the port 4561 is used. 
 
RMR_RTG_ISRAW 
   
  Is set to 1 if the route table generator is sending 
  "plain" messages (not using RMr to send messages, 0 if the 
  rtg is using RMr to send. The default is 1 as we don't 
  expect the rtg to use RMr. 
 
RMR_SEED_RT 
   
  This is used to supply a static route table which can be 
  used for debugging, testing, or if no route table 
  generator process is being used to supply the route table. 
  If not defined, no static table is used and RMr will not 
  report *ready* until a table is received. The static route 
  table may contain both the route table (between newrt 
  start and end records), and the MEID map (between meid_map 
  start and end records) 
 
RMR_SRC_ID 
   
  This is either the name or IP address which is placed into 
  outbound messages as the message source. This will used 
  when an RMR based application uses the rmr_rts_msg() 
  function to return a response to the sender. If not 
  supplied RMR will use the hostname which in some container 
  environments might not be routable. 
 
RMR_VCTL_FILE 
   
  This supplies the name of a verbosity control file. The 
  core RMR functions do not produce messages unless there is 
  a critical failure. However, the route table collection 
  thread, not a part of the main message processing 
  component, can write additional messages to standard 
  error. If this variable is set, RMR will extract the 
  verbosity level for these messages (0 is silent) from the 
  first line of the file. Changes to the file are detected 
  and thus the level can be changed dynamically, however RMR 
  will only suss out this variable during initialisation, so 
  it is impossible to enable verbosity after startup. 
 
RMR_WARNINGS 
   
  If set to 1, RMR will write some warnings which are 
  non-performance impacting. If the variable is not defined, 
  or set to 0, RMR will not write these additional warnings. 
 
 
SEE ALSO 
-------------------------------------------------------------------------------------------- 
 
rmr_alloc_msg(3), rmr_tralloc_msg(3), rmr_call(3), 
rmr_free_msg(3), rmr_init(3), rmr_init_trace(3), 
rmr_get_meid(3), rmr_get_src(3), rmr_get_srcip(3), 
rmr_get_trace(3), rmr_get_trlen(3), rmr_get_xact(3), 
rmr_payload_size(3), rmr_rcv_msg(3), rmr_rcv_specific(3), 
rmr_rts_msg(3), rmr_ready(3), rmr_fib(3), rmr_has_str(3), 
rmr_tokenise(3), rmr_mk_ring(3), rmr_realloc_payload(3), 
rmr_ring_free(3), rmr_set_trace(3), rmr_torcv_msg(3), 
rmr_wh_open(3), rmr_wh_send_msg(3)