Nag 2.x HUP behaviour (potential problem ?)

[Stanley Hopcroft]

Dear Folks,

I am writing to report a potential problem with hangup signal processing 
by Nagios 2.x

With a CVS version (from the last 10 days) of Nag 2.x, the process vsz 
increases by a about 2-5% after a HUP.

Here's the Nag process vsz over the period containing the HUP

Mon May 23 14:20:00 EST 2005,20020
Mon May 23 14:40:01 EST 2005,20028
Mon May 23 15:00:00 EST 2005,20036
Mon May 23 15:20:00 EST 2005,20044
Mon May 23 15:40:01 EST 2005,20044
Mon May 23 16:00:01 EST 2005,20052
Mon May 23 16:20:01 EST 2005,20060
Mon May 23 16:40:00 EST 2005,20068
Mon May 23 17:00:00 EST 2005,20068
Mon May 23 17:20:00 EST 2005,20076
Mon May 23 17:40:01 EST 2005,20076
Mon May 23 18:00:01 EST 2005,20084
Mon May 23 18:20:00 EST 2005,20092
Mon May 23 18:40:00 EST 2005,20092
Mon May 23 19:00:01 EST 2005,20100
Mon May 23 19:20:01 EST 2005,20108
Mon May 23 19:40:01 EST 2005,20108
Mon May 23 20:00:01 EST 2005,20720
Mon May 23 20:20:01 EST 2005,20720
Mon May 23 20:40:01 EST 2005,20720

>From the Nagios log,

Mon May 23 19:42:07 Caught SIGHUP, restarting...
Mon May 23 19:42:07 Nagios 2.0b3 starting... (PID=75013)
Mon May 23 19:42:07 LOG VERSION: 2.0

As you can see, the vsz increases from 20108 kB to 20720 kB in the 
interval containing the HUP (19:40-20:00)

I thought the effect of the HUP would be to free the data structures 
used by the configuration, returning that memory to the process heap. 

Subsequent malloc() to rebuild the configuration data structures woould 
be satisified from the heap, therefore not requiring the process to get 
more memory from the OS and therefore not changing the vsz.

Is this growth in Nagios 2.x vsz normal after a HUP signal ?

(NB this is a Nagios with embedded Perl so there is a greater than usual 
memory leak rate. The surge in memory usage after a HUP is what may be 
unusal).

Yours sincerely.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: disclaimer.txt
URL: <https://www.monitoring-lists.org/archive/users/attachments/20050524/c61d8d16/attachment.txt>