[erlang-questions] Creating a diagnostic memory dump of live erlang VM

[Matthew Evans]

Hi,
I took it on myself to write a little module. This will produce a crash report that will display most of the important data that can be loaded into the crash dump viewer. From Erlang run report:crash_report(). 
The crash dump goes to /tmp/erl_crash.dump and can be viewed from Erlang via your browser when you run crashdump_viewer:start(). from your Erlang shell.
=========

-module(report).
-export([crash_report/0]).
crash_report() ->    Date = erlang:list_to_binary(rfc1123_local_date()),    Header = binary:list_to_bin([<<"=erl_crash_dump:0.2\n">>,Date,<<"\nSystem version: ">>]),    Ets = ets_info(),    Report = binary:list_to_bin([Header,erlang:list_to_binary(erlang:system_info(system_version)),erlang:system_info(info),erlang:system_info(procs),Ets,erlang:system_info(dist),                <<"=loaded_modules\n">>,binary:replace(erlang:system_info(loaded),<<"\n">>,<<"\n=mod:">>,[global])]),    file:write_file("/tmp/erl_crash.dump",Report).
ets_info() ->    binary:list_to_bin([ets_table_info(T)||T<-ets:all()]).
ets_table_info(Table) ->    Info = ets:info(Table),    Owner = erlang:list_to_binary(erlang:pid_to_list(proplists:get_value(owner,Info))),    TableN = erlang:list_to_binary(erlang:atom_to_list(proplists:get_value(name,Info))),    Name = erlang:list_to_binary(erlang:atom_to_list(proplists:get_value(name,Info))),    Objects = erlang:list_to_binary(erlang:integer_to_list(proplists:get_value(size,Info))),    binary:list_to_bin([<<"=ets:">>,Owner,<<"\nTable: ">>,TableN,<<"\nName: ">>,Name,<<"\nObjects: ">>,Objects,<<"\n">>]).
rfc1123_local_date() ->    rfc1123_local_date(os:timestamp()).rfc1123_local_date({A,B,C}) ->    rfc1123_local_date(calendar:now_to_local_time({A,B,C}));rfc1123_local_date({{YYYY,MM,DD},{Hour,Min,Sec}}) ->    DayNumber = calendar:day_of_the_week({YYYY,MM,DD}),    lists:flatten(        io_lib:format("~s, ~2.2.0w ~3.s ~4.4.0w ~2.2.0w:~2.2.0w:~2.2.0w GMT",            [httpd_util:day(DayNumber),DD,httpd_util:month(MM),YYYY,Hour,Min,Sec]));rfc1123_local_date(Epoch) when erlang:is_integer(Epoch) ->    rfc1123_local_date(calendar:gregorian_seconds_to_datetime(Epoch+62167219200)).

From: vladdu55@REDACTED
Date: Thu, 20 Feb 2014 10:30:39 +0100
To: mabrek@REDACTED
CC: erlang-questions@REDACTED
Subject: Re: [erlang-questions] Creating a diagnostic memory dump of live erlang VM

Hi,
On Thu, Feb 20, 2014 at 9:26 AM, Anton Lebedevich <mabrek@REDACTED> wrote:


Another thing which I missed a lot after converting from java to erlang

is a thread dump. It turned out that it's possible to get all

stacktraces for all processes (even with function arguments) via

erlang:system_info(procs).



It returns them as text so it's better to dump it to file immediately:

file:write_file("/tmp/procs.txt",erlang:system_info(procs)).



Format of these traces is quite interesting (undocumented) the best

description I found is in the mailing list

http://erlang.org/pipermail/erlang-questions/2012-November/070609.html
I might just as well ask the obvious question: why is not this information available even as normal Erlang terms, so that one doesn't need to parse it? When writing a crash dump, it doesn't matter, but if it should be used at runtime it's a pain to parse it... This applies to the other results from system_info/1 that are dumped as text.


regards,Vlad


_______________________________________________
erlang-questions mailing list
erlang-questions@REDACTED
http://erlang.org/mailman/listinfo/erlang-questions 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20140220/6e9b9487/attachment.htm>