system.cpu.util not working on HPUX 11.23

[EricWebb]

Hi guys,

Hopefully you can tell me what I'm doing wrong -- I have self-compiled 1.1b5 agentd running on HP-UX 11.23 (1.1b6 server running on Linux). I am trying to add item "system.cpu.util" to my HP-UX host but after I define it and ensure it is set to active, its status is set to "Not supported". I also see this in the log file if I set debug to level 4. Browsing through the code, it looks like it is defined for HP-UX, so I would expect it to work.

I have three items to test various formulations:
system.cpu.util[]
system.cpu.util[all,system,avg1]
system.cpu.util[cpu0,user,avg1]

Errors:
006757:20060306:102511 Got line:system.cpu.util[]
006757:20060306:102511 Sending back:ZBX_NOTSUPPORTED
...
006760:20060306:102527 Got line:system.cpu.util[all,system,avg1]
006760:20060306:102527 Sending back:ZBX_NOTSUPPORTED
...
006758:20060306:102545 Got line:system.cpu.util[cpu0,user,avg1]
006758:20060306:102545 Sending back:ZBX_NOTSUPPORTED

Any ideas?

[EricWebb]

Took a few more minutes to look around... looks like src/libs/zbxsysinfo/hpux/cpu.c calls function get_stat() to read the contents of /tmp/zabbix_agentd.tmp, which should contain stats collected from src/zabbix_agent/stats.c and cpustat.c. The problem is that function collect_stats_cpustat() reads from /proc/stat.

Does HP-UX support the /proc filesystem? I can't find anything on it if it does, and it isn't present on my systems.

Is HP-UX-specific support not yet fully developed? Who from the Zabbix team could I work with to fill in the blanks?


-E.

[fmtaylor2]

HPUX /proc filesystem

According to the information at this link:

http://devresource.hp.com/drc/resources/portguideparisc/index.jsp

No, hpux does not support the /proc filesystem. And since I don't do C there is not much I can do about it, but it seems that they have all the information available to fix the hpux issues in zabbix..........

[Alexei]

I'll be happy to receive a patch supporting monitoring of CPU utilisation under HP-UX. Anyone?

[mariedling]

Here is how to make cpu stats work for HPUX

Here is the diff file for: zabbix-1.1beta9/src/zabbix_agent/cpustat.c

59,63d58
< #ifdef HPUX
< #include <sys/pstat.h>
< #include <sys/dk.h>
< #endif
<
311,319d305
< #ifdef HPUX
< struct pst_dynamic stats;
< float mrtotal;
<
< pstat_getdynamic( &stats, sizeof( struct pst_dynamic ), 1, 0 );
<
< add_values_cpustat(now,stats.psd_cpu_time[CP_USER],stats.psd_cpu_time[CP_SYS],stats.psd_cpu_time[CP_NICE],stats.psd_cpu_time[CP_IDLE]);
<
< #else
340d325
< #endif

And the patch for zabbix-1.1beta9/src/zabbix_agent/stats.c

82d81
< #ifndef HPUX
85d83
< #endif


You also need to define HAVE_PROC_NET_DEV in the config.h file

[brassel]

IS this patch working on zabbix 1.1.1 ?

Hello I am new to zabbix and I try to install the agent on HPUX 11iv1 so I have just installed the pre-compiled agent but I can not graph the cpu usage.

I have seen your patch description but is it working for the 1.1.1 agent version ?

And has someone graphed the cpu use on HPUX ?

Regards

Claude

Here is the diff file for: zabbix-1.1beta9/src/zabbix_agent/cpustat.c

59,63d58
< #ifdef HPUX
< #include <sys/pstat.h>
< #include <sys/dk.h>
< #endif
<
311,319d305
< #ifdef HPUX
< struct pst_dynamic stats;
< float mrtotal;
<
< pstat_getdynamic( &stats, sizeof( struct pst_dynamic ), 1, 0 );
<
< add_values_cpustat(now,stats.psd_cpu_time[CP_USER],stats.psd_cpu_time[CP_SYS],stats.psd_cpu_time[CP_NICE],stats.psd_cpu_time[CP_IDLE]);
<
< #else
340d325
< #endif

And the patch for zabbix-1.1beta9/src/zabbix_agent/stats.c

82d81
< #ifndef HPUX
85d83
< #endif


You also need to define HAVE_PROC_NET_DEV in the config.h file

[den_crane]

Yes, I have graph: system.cpu.load[all,avg15]

$ ./zabbix_agentd -c zabbix_agentd.conf -p |grep cpu
cpu[idle1]                                    [m|ZBX_NOTSUPPORTED]
system.cpu.switches                           [m|ZBX_NOTSUPPORTED]
system.cpu.intr                               [m|ZBX_NOTSUPPORTED]
system.cpu.util[all,user,avg1]                [u|0]
system.cpu.load[all,avg1]                     [d|0.006152]

$ ./zabbix_agentd -c zabbix_agentd.conf -v
ZABBIX Agent (daemon) v1.1.1 (19 July 2006)
Compilation time:  Aug  2 2006 07:39:48

$ uname -a
HP-UX L2000CO2 B.11.11 U 9000/800 1166971547 unlimited-user license

$ ./zabbix_agentd -c /etc/zabbix_agentd.conf -t system.cpu.util[all,system,avg15]
system.cpu.util[all,system,avg15]             [u|0]
$ ./zabbix_agentd -c /etc/zabbix_agentd.conf -t system.cpu.util[all,user,avg15]
system.cpu.util[all,user,avg15]               [u|28]
$ ./zabbix_agentd -c /etc/zabbix_agentd.conf -t system.cpu.util[all,idle,avg15]
system.cpu.util[all,idle,avg15]               [u|58]

[den_crane]

however [util system] - return wrong value

$ ./zabbix_agentd -c /etc/zabbix_agentd.conf -t system.cpu.util[all,system,avg5]
system.cpu.util[all,system,avg5]              [u|0]
$ ./zabbix_agentd -c /etc/zabbix_agentd.conf -t system.cpu.util[all,system,avg15]
system.cpu.util[all,system,avg15]             [u|0]
$ ./zabbix_agentd -c /etc/zabbix_agentd.conf -t system.cpu.util[all,system,avg15]
system.cpu.util[all,system,avg15]             [u|0]

$ sar 1 5

13:57:43    %usr    %sys    %wio   %idle
13:57:44      20      13      10      57
13:57:45      15       3       5      77
13:57:46      17      11      24      48
13:57:47      18      11      23      48
13:57:48      23       7      13      57

Average       19       9      15      57

may be it return not SYS but SSYS?

Load averages: 0.26, 0.31, 0.35
264 processes: 254 sleeping, 10 running
Cpu states:
CPU   LOAD   USER   NICE    SYS   IDLE  BLOCK  SWAIT   INTR   SSYS
 0    0.30  11.9%   0.0%   3.0%  85.1%   0.0%   0.0%   0.0%   0.0%
 1    0.21  17.8%   0.0%   5.9%  76.2%   0.0%   0.0%   0.0%   0.0%
---   ----  -----  -----  -----  -----  -----  -----  -----  -----
avg   0.26  15.0%   0.0%   4.0%  81.0%   0.0%   0.0%   0.0%   0.0%

cpu.util user, cpu.util idle -- return right value.

[arli]

I dug up one old thread, because I believe issue discussed here, is still unresolved.
I can't get the cpu utilization monitoring to work on a multiprocessor HP-UX B.11.31 U ia64 system.
Total utilization displays fine, but if I want to graph it by core, it reports that half of the cpu's are idle, while they are not.
The client side is HP RX6600 with 4 dual core processors.
Both Zabbix Server and Agent are version 1.8.2.

Zabbix server is able to find all 8 cores:

[root@monitoring ~]# zabbix_get -s myserver -k system.cpu.num
8

CPU   LOAD   USER   NICE    SYS   IDLE  BLOCK  SWAIT   INTR   SSYS
 0    1.15  61.6%  20.1%   3.4%  14.9%   0.0%   0.0%   0.0%   0.0%
 2    1.53  75.1%   3.8%   3.4%  17.7%   0.0%   0.0%   0.0%   0.0%
 4    1.52  69.6%  21.7%   3.8%   5.0%   0.0%   0.0%   0.0%   0.0%
 6    1.51  82.9%   6.4%   4.2%   6.6%   0.0%   0.0%   0.0%   0.0%
 8    1.34  82.7%   0.6%   3.6%  13.1%   0.0%   0.0%   0.0%   0.0%
10    1.22  78.1%   6.4%   3.0%  12.5%   0.0%   0.0%   0.0%   0.0%
12    1.23  62.6%  17.3%   4.4%  15.7%   0.0%   0.0%   0.0%   0.0%
14    1.17  54.3%  24.5%   4.2%  17.1%   0.0%   0.0%   0.0%   0.0%
---   ----  -----  -----  -----  -----  -----  -----  -----  -----
avg   1.33  70.8%  12.7%   3.6%  12.9%   0.0%   0.0%   0.0%   0.0%

The utilization numbers here are not exactly the same because of human delay, but the general idea should stand out.

[root@monitoring ~]# zabbix_get -s myserver -k system.cpu.util[0,,]
70.417423
[root@monitoring ~]# zabbix_get -s myserver -k system.cpu.util[1,,]
0.000000
[root@monitoring ~]# zabbix_get -s myserver -k system.cpu.util[2,,]
85.233460
[root@monitoring ~]# zabbix_get -s myserver -k system.cpu.util[3,,]
0.000000
[root@monitoring ~]# zabbix_get -s myserver -k system.cpu.util[4,,]
84.339934
[root@monitoring ~]# zabbix_get -s myserver -k system.cpu.util[5,,]
0.000000
[root@monitoring ~]# zabbix_get -s myserver -k system.cpu.util[6,,]
78.993399
[root@monitoring ~]# zabbix_get -s myserver -k system.cpu.util[7,,]
0.000000

...just in case, I also tried numbers up to 16, but everyting above 7 became ZBX_NOTSUPPORTED
Can someone please try this on HP-UX to confirm this behaviour.