Host based functions

[sauron]

When i'm added in zabbix snmp based host i'm can got snmp based parameters but if host unreachable, i'm need add simple check icmp for got information about this. Zabbix database contain information about this event, but this information can't getted from trigger expression. Why ? Simple, in trigger expression we can use only item based function. And when snmp based host unreachable this item & triggers not working and marked state unknown. I'm think zabbix need host based functions, without item depends. For this devolopers may add specal item system. For example usefull trigger expression:

Check for host unreachable:

{host:system.available(0)}=2

Check for time:

{host:system.time(0)}=101230

and etc.

[tvman]

host based triggers

Hi.
It dosn't make sence. host is point of addressing. host is a place for everything like ip interfaces, services and even OS. for a sample, I use zabbix host for automatic inventory tracking. And I've got a "host" as an address for all my records. Also a host could have more then one interface. How to address it in this case? You have to choose one of interfaces but it's already one of items binded to this host. And how to decide that host is down or not if main interface is down but host still available through another interface? Once again, it doesn't make sence to monitor host instead of items.

[mlevin]

Hi anonymous,

What you ask for (host:system.available), you can do right now with icmpping simple check. Create the item, name it "system", make it the simple check with icmpping. Then, create a trigger for your "system" item, then make your whatever snmp trigger depending on the "system" trigger. Very simple.

Bye.

When i'm added in zabbix snmp based host i'm can got snmp based parameters but if host unreachable, i'm need add simple check icmp for got information about this. Zabbix database contain information about this event, but this information can't getted from trigger expression. Why ? Simple, in trigger expression we can use only item based function. And when snmp based host unreachable this item & triggers not working and marked state unknown. I'm think zabbix need host based functions, without item depends. For this devolopers may add specal item system. For example usefull trigger expression:

Check for host unreachable:

{host:system.available(0)}=2

Check for time:

{host:system.time(0)}=101230

and etc.

[sauron]

What you ask for (host:system.available), you can do right now with icmpping simple check.

He. For example snmp enable host answer only by snmp. If you check this with icmpping you got 0 i.e. unreachable.

Create the item, name it "system", make it the simple check with icmpping. Then, create a trigger for your "system" item, then make your whatever snmp trigger depending on the "system" trigger. Very simple.

I know about this. But it's not need icmp item, need information already contain in database.

[sauron]

It dosn't make sence. host is point of addressing. host is a place for everything like ip interfaces, services and even OS.

it's abstract object contains items with information about interfaces, services and even OS.

for a sample, I use zabbix host for automatic inventory tracking. And I've got a "host" as an address for all my records. Also a host could have more then one interface. How to address it in this case? You have to choose one of interfaces but it's already one of items binded to this host. And how to decide that host is down or not if main interface is down but host still available through another interface? Once again, it doesn't make sence to monitor host instead of items.

Simple. If snmp enable host unreachable trigger based on snmp agent item and this item going to UNKNOWN. Zabbix know about unreachable host, but not process this state in tigger expression. I'm think need virtual functions without item & host solution. My solution may simple added.

[tvman]

Probably I got your idea. You are talking about trigger which fires up when value of the particular item cannot be retrived. Right?

[sauron]

Probably I got your idea. You are talking about trigger which fires up when value of the particular item cannot be retrived. Right?

No Zabbix contain functions independ at items. For example time date & etc. I'm think zabbix need virtual item for support it. It's simple for devolop & understand.

[tvman]

Do you speak in Russian? (according to your location info...)
Sorry, I don't understand your idea. Send me a private message in Russian and I'll try to get what are you talking about.

[sauron]

Do you speak in Russian? (according to your location info...)
Sorry, I don't understand your idea. Send me a private message in Russian and I'll try to get what are you talking about.

Okey.

PS I'm know my english ugly

[James Wells]

Greetings,

I think I understand what you are asking for Sauron.

You are asking for various pseudo item checks that the server executes locally. Things like you showed;

{host:system.time(0)}#{host:local.time(0)}

Where host:local.time(0) is a pseudo item that the server would pretend to check, but instead populates with local data. This would basically trigger if the remote hosts time didn't match the local server's time. Some other items might look like;

• host:local.uname -- To pull the name that the Zabbix server calls the remote system.
• host:local.ipaddr -- To pull the IP address that the Zabbix server sees the remote system as.

Please correct me if I am wrong.

[tvman]

Probably not. Because it's possible right now if snmp agent is capable to provide with this information. If you run zabbix agent on the host you can do whatever you want to. I got message from sauron in Russian but didn't get what he's talking about. I use Zabbix over a year and very familiar with concepts and source code. Let's wait for the next message from him. I think he've got an idea but cannot make it in formal way.

[tvman]

Ok guys, the idea is to expand item key "status" to hosts which have snmp base items only. As I understand item key "statis" based on network_errors field in table hosts. There is only one function get_value_zabbix() which returns NETWORK_ERROR. Counter network_errors increments when function get_value() returns NETWORK_ERROR. All other functions return FAIL. Probably if we would use NETWORK_ERROR instead we would have status key work. I will not test this assumption. If someone want to, please let me know what the result is.

[James Wells]

Ahh, okay, and yes that makes a lot of sense. I will look into this a bit this weekened and see if I can cobble a patch to support this. Please note that if I get it, I will not mainline it, but make it available as a patch only initially. This is one of those items I would need to run by Alexei or Eugene before I try to mainline it.

[tvman]

Cool.
Any way I stuck with my own branch and can do nothing. Maybe I would implement this also. Because it's kind of annoing when one of the the items for some reasons is stop working and there is no any notification about. This "status" key would be useful for indication ANY communication errors.