Upgrade BEAM code cause Mnesia and Ranch crash

[chuanjian666]

We use Mnesia as application's database, when use the cb_admin's upgrade beam code, the mnesia and ranch will be crashed.

11:09:29.166 [error] Supervisor mnesia_kernel_sup had child mnesia_tm started with mnesia_tm:start() at <0.105.0> exit with reason killed in context child_terminated
11:09:29.166 [error] Supervisor mnesia_kernel_sup had child mnesia_tm started with mnesia_tm:start() at <0.105.0> exit with reason reached_max_restart_intensity in context shutdown
11:09:29.167 [error] Supervisor mnesia_sup had child mnesia_kernel_sup started with mnesia_kernel_sup:start() at <0.100.0> exit with reason shutdown in context child_terminated
11:09:29.167 [error] Supervisor mnesia_sup had child mnesia_kernel_sup started with mnesia_kernel_sup:start() at <0.100.0> exit with reason reached_max_restart_intensity in context shutdown
11:09:29.182 [info] Application mnesia exited with reason: shutdown
11:09:29.772 [error] Supervisor {<0.163.0>,ranch_acceptors_sup} had child {acceptor,<0.163.0>,28} started with ranch_acceptor:start_link(#Port<0.22521>, ranch_tcp, <0.162.0>) at <0.191.0> exit with reason killed in context child_terminated
11:09:29.773 [error] Supervisor {<0.163.0>,ranch_acceptors_sup} had child {acceptor,<0.163.0>,29} started with ranch_acceptor:start_link(#Port<0.22521>, ranch_tcp, <0.162.0>) at <0.192.0> exit with reason killed in context child_terminated
11:09:29.773 [error] Supervisor {<0.163.0>,ranch_acceptors_sup} had child {acceptor,<0.163.0>,5} started with ranch_acceptor:start_link(#Port<0.22521>, ranch_tcp, <0.162.0>) at <0.168.0> exit with reason killed in context child_terminated
11:09:29.773 [error] Supervisor {<0.163.0>,ranch_acceptors_sup} had child {acceptor,<0.163.0>,1} started with ranch_acceptor:start_link(#Port<0.22521>, ranch_tcp, <0.162.0>) at <0.164.0> exit with reason killed in context child_terminated
11:09:29.773 [error] Supervisor {<0.163.0>,ranch_acceptors_sup} had child {acceptor,<0.163.0>,2} started with ranch_acceptor:start_link(#Port<0.22521>, ranch_tcp, <0.162.0>) at <0.165.0> exit with reason killed in context child_terminated
11:09:29.773 [error] Supervisor {<0.163.0>,ranch_acceptors_sup} had child {acceptor,<0.163.0>,6} started with ranch_acceptor:start_link(#Port<0.22521>, ranch_tcp, <0.162.0>) at <0.169.0> exit with reason killed in context child_terminated

By reading the mnesia source code, I found this comment in mnesia_sp.erl module

%% To able to generate nice crash reports we need a catch on the highest level.
%% This code can't be purged so a code change is not possible.
%% And hence this a simple module.

So I solved this problem temporarily by filtering mnesia and ranch related modules from code:all_loaded()

Some people also commit this issue in mail list
https://groups.google.com/forum/#!topic/chicagoboss/sKN7Bvpeskc