Re: monit alerts

[Jovan Kostovski]

On 8/14/07, Martin Pala <address@hidden> wrote:
> Currently just global alert ('set alert ...') is implemented this way.

Which means I have to write exec handlers for each monitored process separately.

I want to execute a script whenever process starts, stops, restarts or
encounters some error. The first thing that comes on my mind is to
check the pid file if exists
or it's changed.

To execute the script when process restart I will add in the "check
process" block:
check pidfile with path /path/to/the/pidfile
if changed pid then exec "/path/to/script"

To execute the script when process is stopped:
#check if the pid file does not exist for 2 cycles and execute script
check file pidfile.pid with path /path/to/pid
if failed nonexist for 2 cycles then exec "/path/to/script"

To execute the script when process is started:
# check if the timestamp is less or equal to the monit cycle time
# if the process was started between two cycles
check file pidfile.pid with path /path/to/pid
if timestamp <= <MONIT_CYCLE_TIME> then
exec "/path/to/script"

The problem with this approach (timestamp check) is that
the timestamp will change on process restart (pid change) as well.

How can I check the state of the process for processes that
don't delete their pid file? Is there another way to check the state
of the process except checking its pid file?

How can I create an error handler. Is counting restart atempts good
(if the process can't be restarted x times than an error occured)?
For example:
check file pidfile.pid with path /path/to/pid
if 3 restarts in 3 cycles then exec "/path/to/script"

BR, Jovan