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

[metaworx lists]

> On Behalf Of Jan-Henrik Haukeland
> Sent: Dienstag, 20. Dezember 2005 17:46
>
> On 20. des. 2005, at 11.25, metaworx lists wrote:
> > *
> 
> Thanks for the trace and valgrind output. I could not see anything  
> directly wrong, but I'm pretty sure what the problem is now.

wellcome. I think: of course isn't anything wrong, since it works as long as
I use strace or valgrind. :-)

the problem only occures, if I run plain monit. this talks against your
explanation. because then it would not work even if I use strace.

> > ========> here i called sudo monit -Iv stop vs5_dbhost
> 
> I think this may be the problem; when you call stop on a 
> program, the  
> monit damon thread which runs the http server will stop handling new  
> incoming calls until the program is actually stopped. (The socket  
> connection from the monit client is accepted by the kernel so this  
> works, but it is ignored by the monit server). This problem is most  
> notable when the program to be stopped takes a long time to stop.  
> This is an old problem, not particular to the new beta. We do 
> plan to  
> fix this, see http://www.tildeslash.com/monit/doc/next.php#23
> 
> To test if this is the case, instead of stop try this
> 
> 1) start monit in debug mode, monit -Iv
> 2) toggle unmonitor and monitor on vs5_dbhost by, monit -v unmonitor  
> vs5_dbhost and monit -v monitor vs5_dbhost

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

> 3) Do you see this in the console:
>       Monitoring disabled -- service vs5_dbhost
>       Monitoring enabled -- service vs5_dbhost

see above. (yes with strace/valgrind, no without)

> 4) If yes, then your problem is connected to the stop "design bug"  
> explained above.

unfortunately it seems not to be. and in v4.6 and v4.5.1 it works perfectly
with the very same config.

> 5) If no, let us know.

herewith I do. (due to the delay of the mailinglist (4h) I send it bcc to
jan-henrik and martin p.).

martin r.