trigger not working, but expression looks right

[groknaut]

hello,

could someone help me with this trigger? this trigger compares current JVM memory usage (RSS) with the current total memory on the system. if the RSS memory of the JVM is greater than 70% of total memory, it should trigger an average severity alert. but it's not triggering.

OS: CentOS 5.7
zabbix: 1.8.10

here's the expression:

{Template_Java_OS_Resources:check_health[mem,java,rss].last(0)}>({Template_Linux:vm.memory.size[total].last(0)}*0.7)

check_health is a script on the system, called as a UserParameter in zabbix-agent.conf.

UserParameter=check_health[*],sudo /usr/local/bin/check_health.rb $1 $2 $3

we do indeed get data from this check:

webapp# su - zabbix
webapp$ sudo check_health.rb mem java rss
5541860

and given the total memory on the system,

webapp$ free | grep ^Mem | awk '{print $2}'
7347752

that trigger should be engaged, as 5541860 is greater than 70% of total memory.

thanks,
kallen

[JBo]

Hi,

Are you getting these data in Zabbix ?
You can check them in Monitoring / Latest data.

Regards,
JBo

[heaje]

Another thing to check is to be sure you're comparing similar units. For instance, it looks like your check_health[mem,java,rss] item is collecting data in kilobytes. However, if my memory serves me well, vm.memory.size[total] is in Bytes. Therefore, you would have to be using a LOT of kbytes of memory to be more than 70% of your memory in Bytes .

A simple thing to do would be to multiply the value of check_health[mem,java,rss] by 1024 to get the value in bytes (assuming it is in kilobyes).

[groknaut]

yep, i'm definately getting the data in zabbix for both measurements.

and AHA! units! i just fixed that bit by changing how check_health reports the data. it now reports it in bytes, which is how vm.memory.size[total] is measured.

but my trigger still isn't firing. i've even changed it to fire at 20%, but no joy.
any help?

trigger expression:

{Template_Java_OS_Resources:check_health[mem,java,rss].last(0)}>({Template_Linux:vm.memory.size[total].last(0)}*0.2)

right now, java pid's RSS is 63% of total mem:

4749471744 / 7524098048 = .63123469599953836220

evidence on the wire of what's being sent:

[root@webapp ~]# tcpflow -c port 10050 | egrep -A 4 'vm.memory.size\[total\]|check_health'
tcpflow[21279]: listening on eth0
010.0xx.038.143.50816-010.192.103.031.10050: check_health[mem,java,rss]

010.1xx.013.031.10050-010.0xx.018.143.50816: ZBXD.
010.1xx.013.031.10050-010.0xx.018.143.50816:
.......4749471744
--
010.0xx.018.143.54239-010.1xx.013.031.10050: vm.memory.size[total]

010.1xx.103.031.10050-010.0xx.018.143.54239: ZBXD.
010.1xx.103.031.10050-010.0xx.018.143.54239:
.......7524098048

[heaje]

Your check looks good based on what you posted. The last thing I can think to try is to modify your trigger slightly. Try changing it to look like this (it accomplishes the same thing):

100*({Template_Java_OS_Resources:check_health[mem,java,rss].last(0)}/{Template_Linux:vm.memory.size[total].last(0)})>20

I know it's just another way of calculating the same thing, but I have many checks that use that exact same logic for triggers (and they work fine).

[groknaut]

i tried what you suggested, but that too is not working.

anyone have more ideas?

is there a server debugging level that would expose how this logic is playing out? i don't think the standard debug level on the server shows me this.

[groknaut]

still not working. and i wonder if i'm running into this unsolved bug:

https://support.zabbix.com/browse/ZBX-4575