Re: 4.7-beta1: monit <command> <service> does not work

[Jan-Henrik Haukeland]

On 20. des. 2005, at 19.52, metaworx lists wrote:

> I tried that. works with strace, doesn't without.

Bhaa! Well, the only (non-commercial[1]) option left then, since we  
cannot reproduce it, is to run gdb and debug the problem.

I'll give you a few pointers to get you started. First, run gdb on  
the non-installed monit software (i.e. from the directory you built  
monit), because debug info is stripped during installation. The idea  
is to test if the monit daemon gets a connection from the monit  
client at all, which seems to be the problem and then find where the  
problem is located. The monit client-server system looks like this:

(1)monit <service> <command>  - (2)send command via socket ->  (3) 
monit daemon accepts command -> (4)do command


1) Open two consoles
2) Start the monit daemon using gdb and set a breakpoint like so;

gdb ./monit
...
(gdb) b socket_create_a
Breakpoint 1 at 0xd8ec: file socket.c, line 256.
(gdb) r -Iv
...

3) In the other console run, 'monit stop vs5_dbhost'

4) If console number 1 stops execution with this;

Breakpoint 1, socket_create_a (socket=6, remote_host=0xa000ce68  
"127.0.0.1", port=2812, sslserver=0x0) at socket.c:256
256       ASSERT(socket>=0);
(gdb) n

Then the connection was accepted by the daemon, now step through the  
code using 'n' or 's' where applicable and try to locate the problem.  
It may be a SSL problem and I'm particularly interested in if this  
block of code in socket.c:socket_create_a,  'if(sslserver)..' has any  
problems? If so, have you recently installed a new open-ssl version?


5) if it does not stop in 4) the problem must be at the other end.  
Instead of setting the breakpoint in the server, start the client  
with gdb and set a breakpoint there as shown below. Make sure you  
have started the daemon first with ./monit -Iv

gdb ./monit
...
(gdb) b control_service_daemon
Breakpoint 1 at 0x39d8: file control.c, line 92.
(gdb) r monitor apache
Starting program: /Users/hauk/src/monit/monit monitor apache
...
Breakpoint 1, control_service_daemon (S=0xbffffa38 "apache",  
action=0xbffffa30 "monitor") at control.c:92
92        char *auth= Util_getBasicAuthHeader();

Step through and see where and why the socket calls fails.

Let us know how it goes!

--
Jan-Henrik Haukeland
Mobil +47 97141255

[1] If all else fails we do provide online help for a small fee, see  
the monit support pages.