Greetings, all;
I'm posting this here to see if anyone wants to comment. Otherwise I'd probably just list it in the issue tracker. Also, I can't seem to reach the support site from here at the moment, so I can't just put in a request there.
I think Zabbix needs a better system for A) Unique identifiers for targets with agents and B) A better system for showing whether an agent is correctly configured in the GUI.
We're discovering here that despite being told many times that the name of the host in the agent config file and the label on the GUI have to match, our system admins have failed to do so repeatedly. This is mostly because they can add the host with the wrong name and still have it show up with a green icon in the "Availability" column of Configuration/Hosts. Some? Of the checks seem to work at least partially even with this problem.
There are currently 500+ servers being tracked with a mix of agents, pings, and SNMP that show green there, yet are still putting entries in our log repeatedly saying "Sending list of active checks to XXX failed, not found".
The status of these icons in the GUI does not change when it should in some cases, like when an agent is switched from active to passive because it's behind a firewall, and they stay red forever even though the agent is working and we are receiving the passive checks (a misnomer, too... the opposite of the name I'd expect active vs. passive).
Also, the need to match host name on the agent with the GUI icon name is an issue... our admins want to name icons with something that makes sense to them, and they assume that a valid host name or FQDN will match because it's "valid" for DNS even though Zabbix requires a string match.
Could you maybe separate the icon label from the unique key that identifies the target? That plus a clear sign that there's a mismatch in the name somewhere would help a lot.
Another thought - the inventory data items don't get much use here because they don't match our CMDB items. What we need is a way to define a tree of arbitrary attributes with data types for a monitored target that we can fill via editing, mass change, or XML import. That way we can add just the data items we need for system contact, contact phone, service contract, serial number, etc.
We'd need to be able to access the attributes for rules checks for triggers so we can alter trigger behavior based on the attributes.. that way we can for example set up a string attribute to say "production" or "dev" and use a unified set of triggers/thresholds for all the systems associated with one application.
Speaking off applications, a more generic and flexible method for combining monitored targets into synthetic objects would be great, so we can pick and choose items to combine into an application or target roll-up. For instance, we could pick the status of the power and temperature items in one data center room and combine with the system and application status to get a red/yellow/green sort of overall view of an application to present to management.
I'll put some more suggestions in the support site when I can get back into it.
Erik
University of Minnesota
PS: To be honest, we here have the necessary time free to do some GUI coding and maybe work with the back end a little too, but we can't find a good reference to Zabbix internals for programmers. Does anyone outside the Zabbix company do any non trivial development for this system?
I'm posting this here to see if anyone wants to comment. Otherwise I'd probably just list it in the issue tracker. Also, I can't seem to reach the support site from here at the moment, so I can't just put in a request there.
I think Zabbix needs a better system for A) Unique identifiers for targets with agents and B) A better system for showing whether an agent is correctly configured in the GUI.
We're discovering here that despite being told many times that the name of the host in the agent config file and the label on the GUI have to match, our system admins have failed to do so repeatedly. This is mostly because they can add the host with the wrong name and still have it show up with a green icon in the "Availability" column of Configuration/Hosts. Some? Of the checks seem to work at least partially even with this problem.
There are currently 500+ servers being tracked with a mix of agents, pings, and SNMP that show green there, yet are still putting entries in our log repeatedly saying "Sending list of active checks to XXX failed, not found".
The status of these icons in the GUI does not change when it should in some cases, like when an agent is switched from active to passive because it's behind a firewall, and they stay red forever even though the agent is working and we are receiving the passive checks (a misnomer, too... the opposite of the name I'd expect active vs. passive).
Also, the need to match host name on the agent with the GUI icon name is an issue... our admins want to name icons with something that makes sense to them, and they assume that a valid host name or FQDN will match because it's "valid" for DNS even though Zabbix requires a string match.
Could you maybe separate the icon label from the unique key that identifies the target? That plus a clear sign that there's a mismatch in the name somewhere would help a lot.
Another thought - the inventory data items don't get much use here because they don't match our CMDB items. What we need is a way to define a tree of arbitrary attributes with data types for a monitored target that we can fill via editing, mass change, or XML import. That way we can add just the data items we need for system contact, contact phone, service contract, serial number, etc.
We'd need to be able to access the attributes for rules checks for triggers so we can alter trigger behavior based on the attributes.. that way we can for example set up a string attribute to say "production" or "dev" and use a unified set of triggers/thresholds for all the systems associated with one application.
Speaking off applications, a more generic and flexible method for combining monitored targets into synthetic objects would be great, so we can pick and choose items to combine into an application or target roll-up. For instance, we could pick the status of the power and temperature items in one data center room and combine with the system and application status to get a red/yellow/green sort of overall view of an application to present to management.
I'll put some more suggestions in the support site when I can get back into it.
Erik
University of Minnesota
PS: To be honest, we here have the necessary time free to do some GUI coding and maybe work with the back end a little too, but we can't find a good reference to Zabbix internals for programmers. Does anyone outside the Zabbix company do any non trivial development for this system?
Comment