Readme in Chinese 中文
molten is transparency tool for application tracing it self module call.
It trace php app core call and output zipkin/opentracing format trace log.
Provides features about muliti trace sapi, multi sampling type, upload tracing status, module control and muliti sink type.
It very easy to build a distributed systems tracing infrastructure base on php, already run on thousand instance on production env.
Everything you should need to install molten on your system.
phpize
./configure
make && make install
make install
copies molten.so
to and appropriate location, but you still need to enable the module int the PHP config file, To do so, either edit your php.ini or add ad molten.ini file in /etc/php5/conf.d
with the following contents:extension=molten.so
.
at ./configure
phase, you can also add --enable-zipkin-header=yes
to support zipkin B3 header.
cd example
sh run.sh
open http://127.0.0.1:9411/zipkin/
in your browser, you can see the tracing detail in it.
if you think above is too simple, you also can do this.
cd example
sh complex.sh
it is cool, alright?
NOTICE
if you not see the trace, you select EndTime +1 hour.
molten.enable
set 1 enable or set 0 disable, default 1
.
molten.service_name
is to set service name for label app service name.
molten.tracing_cli
set 1 to trace, cli sapi, 0 not trace, just use for trace unit test, default 0
.
molten.sampling_type
sampling type choose to use sampling by rate(1
) or request(2
), default is 1
.
molten.sampling_request
sampling by request, set to per min request num, defualt is 1000
.
molten.sampling_rate
determine a request sampled or not by rate, default is 64
.
molten.notify_uri
the uri for molten to notify manger.
report module output type is same as sink module
molten.report_interval
reporter call interval, default 60
.
molten.report_limit
reporter list limit current only for error num, default 100
.
molten.sink_type
sink type, 1
write log, log path is depend on molten.sink_log_path
, 2
write log to standand output, 3
write log to syslog, 4
use curl to send trace log, http uri depend on molten.sink_http_uri
.
molten.output_type
output spans on one line(1
) or one line one span(2
).
molten.sink_log_path
locate log path.
molten.sink_http_uri
locate log http uri.
molten.sink_syslog_unix_socket
transform log to syslog udp unix domain collector.
molten.span_format
span format, you can select zipkin
or opentracing
for different tracing system.
php -d extension=molten.so -d molten.enable=1 -d molten.sink_type=2 -d molten.tracing_cli=1 -d molten.sampling_rate=1 -r '$c=curl_init("http://localhost:12345");curl_exec($c);'
You can see output below:
[{"traceId":"%s","name":"php_curl","version":"php-4","id":"1.1","parentId":"1","timestamp":%d,"duration":%d,"annotations":[{"value":"cs","timestamp":%d,"endpoint":{"serviceName":"%s","ipv4":"%s"}},{"value":"cr","timestamp":%d,"endpoint":{"serviceName":"%s","ipv4":"%s"}}],"binaryAnnotations":[{"key":"http.url","value":"http:\/\/localhost:12345\/","endpoint":{"serviceName":"%s","ipv4":"%s"}},{"key":"error","value":"Failed
connect to localhost:12345; Connection
refused","endpoint":{"serviceName":"%s","ipv4":"%s"}}]},{"traceId":"%s","name":"cli","version":"php-4","id":"1","timestamp":%d,"duration":%d,"annotations":[{"value":"sr","timestamp":%d,"endpoint":{"serviceName":"%s","ipv4":"%s"}},{"value":"ss","timestamp":%d,"endpoint":{"serviceName":"%s","ipv4":"%s"}}],"binaryAnnotations":[{"key":"path","value":"-","endpoint":{"serviceName":"%s","ipv4":"%s"}}]}]
the Config Block above, you can see which feature we support.
we intercept curl,pdo,mysqli,redis,mongodb,memcached extension to build execute time span info. for chain http request, we replace curl_exec,curl_setopt,curl_setopt_array to add http request trace header (x-w-traceid, x-w-spanid and so on).
the span_format is the way to custom span format, for two popular kinds (zipkin
and opentracing
).
different sampling type and change parameter to control sampling, rate or request.
Sink is the output where you locate, molten support to standard fd, file, http and others (continue), on this way, we can choose where to output trace log.
Use http to control our sampling.
see molten status, request http://domain/molten/status
use GET method.
the output is below, already adapt the style of prometheus.
# HELP molten_request_all Number of all request.
# TYPE molten_request_all counter
molten_request_all %d
# HELP molten_request_capture Number of request be capture.
# TYPE molten_request_capture counter
molten_request_capture %d
# HELP molten_sampling_type the type of sampling.
# TYPE molten_sampling_type gauge
molten_sampling_type %d
# HELP molten_sampling_rate the rate of sampling.
# TYPE molten_sampling_rate gauge
molten_sampling_rate %d
# HELP molten_sampling_request the request be capture one min.
# TYPE molten_sampling_request gauge
molten_sampling_request %d
modify molten sampling, request http://domain/molten/status
use POST method,
body is json format, field has the same meaning of config.
{"enable":1,"samplingType":2,"samplingRate":20,"samplingRequest":100}
Report base import info which we do not sampled like error list.
Welcome developers who willing to make it better.
the mail list below you can contract for discuss and improve more power.
You may contribute in the following ways:
- Repost issues and feedback.
- Submit fixes, features via Pull Request.
Apache License Version 2.0 see http://www.apache.org/licenses/LICENSE-2.0.html