haproxy/contrib/opentracing/README-func
Miroslav Zagorac 70230c6484 CONTRIB: opentracing: add the OpenTracing filter
This commit adds the OpenTracing filter (hereinafter we will use the
abbreviated name 'the OT filter') to the contrib tree.

The OT filter adds native support for using distributed tracing in HAProxy.
This is enabled by sending an OpenTracing compliant request to one of the
supported tracers; such as Datadog, Jaeger, Lightstep and Zipkin tracers.
Please note: tracers are not listed by any preference, but alphabetically.

The OT filter is a standard HAProxy filter, so what applies to others also
applies to this one (of course, by that I mean what is described in the
documentation, more precisely in the doc/internals/filters.txt file).

The OT filter activation is done explicitly by specifying it in the HAProxy
configuration.  If this is not done, the OT filter in no way participates
in the work of HAProxy.

As for the impact on HAProxy speed, this is documented with several tests
located in the test directory, and the result is found in the README-speed-*
files.  In short, the speed of operation depends on the way it is used and
the complexity of the configuration, from an almost immeasurable impact to
a significant deceleration (5x and more).  I think that in some normal use
the speed of HAProxy with the filter on will be quite satisfactory with a
slowdown of less than 4%.

The OT filter allows intensive use of ACLs, which can be defined anywhere in
the configuration.  Thus, it is possible to use the filter only for those
connections that are of interest to us.

More detailed documentation related to the operation, configuration and use
of the filter can be found in the contrib/opentracing directory.

To make the OpenTracing filter easier to configure and compile, several
entries have been added to the Makefile.  When running the make utility,
it is possible to use several new arguments:

  USE_OT=1     : enable the OpenTracing filter
  OT_DEBUG=1   : compile the OpenTracing filter in debug mode
  OT_INC=path  : force the include path to libopentracing-c-wrapper
  OT_LIB=path  : force the lib path to libopentracing-c-wrapper
  OT_RUNPATH=1 : add libopentracing-c-wrapper RUNPATH to haproxy executable

If USE_OT is set, then an additional Makefile from the contrib/opentracing
directory is included in the compilation process.
2020-12-16 15:49:53 +01:00

299 lines
9.0 KiB
Plaintext

Here I will write down some specifics of certain parts of the source, these are
just some of my thoughts and clues and they are probably not too important for
a wider audience.
src/parser.c
------------------------------------------------------------------------------
The first thing to run when starting the HAProxy is the flt_ot_parse() function
which actually parses the filter configuration.
In case of correct configuration, the function returns ERR_NONE (or 0), while
in case of incorrect configuration it returns the combination of ERR_* flags
(ERR_NONE here does not belong to that bit combination because its value is 0).
One of the parameters of the function is <char **err> in which an error message
can be returned, if it exists. In that case the return value of the function
should have some of the ERR_* flags set.
Let's look at an example of the following filter configuration what the function
call sequence looks like.
Filter configuration line:
filter opentracing [id <id>] config <file>
Function call sequence:
flt_ot_parse(<err>) {
/* Initialization of the filter configuration data. */
flt_ot_conf_init() {
}
/* Setting the filter name. */
flt_ot_parse_keyword(<err>) {
flt_ot_parse_strdup(<err>) {
}
}
/* Setting the filter configuration file name. */
flt_ot_parse_keyword(<err>) {
flt_ot_parse_strdup(<err>) {
}
}
/* Checking the configuration of the filter. */
flt_ot_parse_cfg(<err>) {
flt_ot_parse_cfg_tracer() {
}
...
flt_ot_post_parse_cfg_tracer() {
}
flt_ot_parse_cfg_group() {
}
...
flt_ot_post_parse_cfg_group() {
}
flt_ot_parse_cfg_scope() {
}
...
flt_ot_post_parse_cfg_scope() {
}
}
}
Checking the filter configuration is actually much more complicated, only the
name of the main function flt_ot_parse_cfg() that does it is listed here.
All functions that use the <err> parameter should set the error status using
that pointer. All other functions (actually these are all functions called
by the flt_ot_parse_cfg() function) should set the error message using the
ha_warning()/ha_alert() HAProxy functions. Of course, the return value (the
mentioned combination of ERR_* bits) is set in all these functions and it
indicates whether the filter configuration is correct or not.
src/group.c
------------------------------------------------------------------------------
The OT filter allows the use of groups within which one or more 'ot-scope'
declarations can be found. These groups can be used using several HAProxy
rules, more precisely 'http-request', 'http-response', 'tcp-request',
'tcp-response' and 'http-after-response' rules.
Configuration example for the specified rules:
<rule> ot-group <filter-id> <group-name> [ { if | unless } <condition> ]
Parsing each of these rules is performed by the flt_ot_group_parse() function.
After parsing the configuration, its verification is performed via the
flt_ot_group_check() function. One parsing function and one configuration
check function are called for each defined rule.
flt_ot_group_parse(<err>) {
}
...
flt_ot_group_check() {
}
...
When deinitializing the module, the function flt_ot_group_release() is called
(which is actually an release_ptr callback function from one of the above
rules). One callback function is called for each defined rule.
flt_ot_group_release() {
}
...
src/filter.c
------------------------------------------------------------------------------
After parsing and checking the configuration, the flt_ot_check() function is
called which associates the 'ot-group' and 'ot-scope' definitions with their
declarations. This procedure concludes the configuration of the OT filter and
after that its initialization is possible.
flt_ops.check = flt_ot_check;
flt_ot_check() {
}
The initialization of the OT filter is done via the flt_ot_init() callback
function. In this function the OpenTracing API library is also initialized.
It is also possible to initialize for each thread individually, but nothing
is being done here for now.
flt_ops.init = flt_ot_init;
flt_ot_init() {
flt_ot_cli_init() {
}
/* Initialization of the OpenTracing API. */
ot_init(<err>) {
}
}
flt_ops.init_per_thread = flt_ot_init_per_thread;
flt_ot_init_per_thread() {
}
...
After the filter instance is created and attached to the stream, the
flt_ot_attach() function is called. In this function a new OT runtime
context is created, and flags are set that define which analyzers are used.
flt_ops.attach = flt_ot_attach;
flt_ot_attach() {
/* In case OT is disabled, nothing is done on this stream further. */
flt_ot_runtime_context_init(<err>) {
flt_ot_pool_alloc() {
}
/* Initializing and setting the variable 'sess.ot.uuid'. */
if (flt_ot_var_register(<err>) != -1) {
flt_ot_var_set(<err>) {
}
}
}
}
When a stream is started, this function is called. At the moment, nothing
is being done in it.
flt_ops.stream_start = flt_ot_stream_start;
flt_ot_stream_start() {
}
Channel analyzers are called when executing individual filter events.
For each of the four analyzer functions, the events associated with them
are listed.
Events:
- 1 'on-client-session-start'
- 15 'on-server-session-start'
------------------------------------------------------------------------
flt_ops.channel_start_analyze = flt_ot_channel_start_analyze;
flt_ot_channel_start_analyze() {
flt_ot_event_run() {
/* Run event. */
flt_ot_scope_run() {
/* Processing of all ot-scopes defined for the current event. */
}
}
}
Events:
- 2 'on-frontend-tcp-request'
- 4 'on-http-body-request'
- 5 'on-frontend-http-request'
- 6 'on-switching-rules-request'
- 7 'on-backend-tcp-request'
- 8 'on-backend-http-request'
- 9 'on-process-server-rules-request'
- 10 'on-http-process-request'
- 11 'on-tcp-rdp-cookie-request'
- 12 'on-process-sticking-rules-request
- 16 'on-tcp-response'
- 18 'on-process-store-rules-response'
- 19 'on-http-response'
------------------------------------------------------------------------
flt_ops.channel_pre_analyze = flt_ot_channel_pre_analyze;
flt_ot_channel_pre_analyze() {
flt_ot_event_run() {
/* Run event. */
flt_ot_scope_run() {
/* Processing of all ot-scopes defined for the current event. */
}
}
}
Events:
- 3 'on-http-wait-request'
- 17 'on-http-wait-response'
------------------------------------------------------------------------
flt_ops.channel_post_analyze = flt_ot_channel_post_analyze;
flt_ot_channel_post_analyze() {
flt_ot_event_run() {
/* Run event. */
flt_ot_scope_run() {
/* Processing of all ot-scopes defined for the current event. */
}
}
}
Events:
- 13 'on-client-session-end'
- 14 'on-server-unavailable'
- 20 'on-server-session-end'
------------------------------------------------------------------------
flt_ops.channel_end_analyze = flt_ot_channel_end_analyze;
flt_ot_channel_end_analyze() {
flt_ot_event_run() {
/* Run event. */
flt_ot_scope_run() {
/* Processing of all ot-scopes defined for the current event. */
}
}
/* In case the backend server does not work, event 'on-server-unavailable'
is called here before event 'on-client-session-end'. */
if ('on-server-unavailable') {
flt_ot_event_run() {
/* Run event. */
flt_ot_scope_run() {
/* Processing of all ot-scopes defined for the current event. */
}
}
}
}
After the stream has stopped, this function is called. At the moment, nothing
is being done in it.
flt_ops.stream_stop = flt_ot_stream_stop;
flt_ot_stream_stop() {
}
Then, before the instance filter is detached from the stream, the following
function is called. It deallocates the runtime context of the OT filter.
flt_ops.detach = flt_ot_detach;
flt_ot_detach() {
flt_ot_runtime_context_free() {
flt_ot_pool_free() {
}
}
}
Module deinitialization begins with deinitialization of individual threads
(as many threads as configured for the HAProxy process). Because nothing
special is connected to the process threads, nothing is done in this function.
flt_ops.deinit_per_thread = flt_ot_deinit_per_thread;
flt_ot_deinit_per_thread() {
}
...
For this function see the above description related to the src/group.c file.
flt_ot_group_release() {
}
...
Module deinitialization ends with the flt_ot_deinit() function, in which all
memory occupied by module operation (and OpenTracing API operation, of course)
is freed.
flt_ops.deinit = flt_ot_deinit;
flt_ot_deinit() {
ot_close() {
}
flt_ot_conf_free() {
}
}