[erlang-questions] Erlang tracing

ILYA Khlopotov <>
Wed Sep 23 17:42:34 CEST 2015


+1 for multiple independent tracers.

It would be also very helpful to support following:
1. Cluster avare seq_trace. Currently trace token is not send over erlang
distribution protocol. It could be solved if new control message is added
to distribution protocol for example.
2. There are a lot of different formats for trace messages and any tracing
library need to be able to understand all of them. Almost any tracing
librbary has to do something like:
https://github.com/erlang/otp/blob/a2670f0822fc6729df956c8ec8c381340ff0a5fb/lib/et/src/et_selector.erl#L218
https://github.com/erlang/otp/blob/a2670f0822fc6729df956c8ec8c381340ff0a5fb/lib/et/src/et_selector.erl#L268
https://github.com/erlang/otp/blob/a2670f0822fc6729df956c8ec8c381340ff0a5fb/lib/runtime_tools/src/dbg.erl#L942
https://github.com/ferd/recon/blob/master/src/recon_trace.erl#L482
It is problematic though since tuples are of different size.
Would it be better to send this info as an erlang record which anyone could
import and use? It could be two records as well. Since every trace message
has a meta info which describes where and what happend and list of
arguments of the message. I.e. something like
meta: {
   pid,
   ts,
   type,
   serial,
   ctx
}

event: {
   meta :: #meta{},
   args :: [],
   sequence = 0
}

3. For seq_trace in particular it could be useful to be able to inject some
context into the tracing process. Consider following use case. Let's say we
have a notion of sessions. First user logins in order to start session,
then he issues some commands using session_cookie as a mean to prove his
identity. Since calls to login and command are independent http requests
there is no way to link them together. We could have following:
maybe_trace_login(_User, _Cookie) ->
    ok. %% function we use for dbg:fun2ms
login(User, Password) ->
   %% check password raise if it is invalid
   Cookie = ...,
   maybe_trace_login(User, Cookie),
   ...
command(Cookie, Command, Args) ->
   %% check Cookie raise if unknown
   do_command(Command, Args), %% we want to get trace of this
   ...
However if we could store {User, Cookie} in the context of a tracer process
from match spec.
We could produce nice trace messages for calls to command which would
include user name.
This is the reason why ctx field is added in #meta{}.
We also would need to be able to initiate tracing of command when cookie
match somehow.

BR,
ILYA

On Mon, Sep 21, 2015 at 1:52 AM, Lukas Larsson <> wrote:

> Hello everyone.
>
> As you may know, one of the OTP teams focus areas for the coming year is
> make tracing better. At the moment we are gathering ideas and attempting to
> put together a vision of what we would like to have, before deciding what
> we can make.
>
> I'm pretty sure that many of you have much more experience with using
> Erlang tracing while developing and in production than I do, which is we
> would love to have your input as to what you would like to change about
> tracing.
>
> To set the scope of the discussion, when I say tracing I include; erlang
> tracing, dtrace/systemtap, trace outputs (stdout/file/IP), filtering
> through match specs, sequence tracing, tool integration (dbg, fprof,
> redbug, recon to mention some) and probably more.
>
> To start the discussion, here are a few of my ideas in no particular order:
>   * Allow multiple tracers. Today only one port/process can be the
> receiver of trace data.
>   * Create a couple of scalable high throughput tracing output backends
> with different overflow mechanics. Today all tracing is funneled through
> one bottleneck and has no overflow handling at all.
>   * Expose vm probes (today dtrace probes) to the erlang tracer.
>   * Better integration of dtrace/lttngt/systemtap into the erlang trace.
>   * Allow the erlang tracer to be an Erlang callback module. Today only
> ports/processes are allowed.
>   * Optimize trace output to file/ip. Maybe use something like the Common
> Trace Format (
> http://git.efficios.com/?p=ctf.git;a=blob_plain;f=common-trace-format-specification.md;hb=master),
> instead of the term_to_binary that we have today.
>   * Write much much better documentation for dbg :)
>
> We are looking for feedback from both beginners as well as seasoned
> veterans to make erlang tracing the best it can be. So if you have any
> thoughts or ideas, join the discussion to make Erlang tracing better for
> you and everyone else.
>
> Thanks in advance,
> Lukas
> Erlang/OTP team
>
> _______________________________________________
> erlang-questions mailing list
> 
> http://erlang.org/mailman/listinfo/erlang-questions
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20150923/b1e0f38c/attachment.html>


More information about the erlang-questions mailing list