What are the advantages/disadvantages to each? The manual does not really explain (at all) about Active checks and the example templates do not use them.
-
-
I would say that the biggest difference, is who is contacting who. This is important if you are working with servers behind NAT, or where you don't have possibility to open the firewall to a listening port.
As far as I know there are no difference in performance/functionality between the two modes. -
Does the active check not mean that the agent buffers its results if the server is unavailable?Comment
-
Comment
-
I just found this page in the wiki which goes in depth with the differences: http://www.zabbix.com/wiki/doku.php?...strapper:index
There is also documentation on the protocol used:
http://www.zabbix.com/wiki/doku.php?...xagentprotocolComment
-
I personally only use active checks because all but maybe one of my agents are behind firewalls. The only limitation to active checking agents (that I can think of) is that actions can not run commands on the agent because Zabbix requires that the server make a connection to the agent, which is not possible if its behind a firewall like in my case.
I also hear that performance is better on the Zabbix server because it doesn't have to ask the agents for their metrics; The agents just send it.
The above is for 1.4.x. There may be more improvements with 1.6, which I haven't touched yet.Comment
-
If you have lots of servers (hundreds+), active checks is the way to go. This will decrease the workload of the zabbix server and shorted the queue's so you wont miss poll intervals.
This is really important if you also have snmp checks, because normal agent checks are done by the same server processes which do snmp checks.
Also nice for testing purposes; if you restart an agent which is configured for active checks, it will send data for all checks regardless of its interval (to create a baseline). If you have hourly+ checks, this is great versus the normal agent checks, with which you'd have to wait up to an hour+.Comment
-
Hmm the downside I am seeing here is increased load on the agent and thus the server you are monitoring. I'd rather the Zabbix server deal with monitoring load rather than say our web/mail/dns/radius etc.. servers.Comment
-
Increased load?! I don't think you're serious here.
Use of active checks decreases (yes, decreases) load on both server and agent sides!Comment
-
Wrong logic detected! Yes, it can disappear. Think about it, ZABBIX server doesn't have to ask agents anymore. The functionality is not needed. No one has to take the load, the load is gone.Comment
-
- checks are done regardless of active / passive
- in passive mode, the server demands and the client executes (can be all checks within 0.1 sec), active mode can queue this itself (dont know if it does tho)
- Active checks actually generate less network traffic
- Negotiation of checks is done at startup and some larger interval (communicating what the active agent should do and when), this is negligible.
- the queuing, ofcourse, requires no more resources than running a normal agent. These are sleeps / waits.Comment
-
I am trying to test active checks on one of my hosts, specifically as I am trying to do log monitoring. However no data seems to be being sent to my zabbix server.
It can definitely connect (checked with telnet). Is there actually any documentation on this feature that can help me? Is there something special I need to do in either the agent config or server config.
Thanks.Comment
I am just saying that if it decreased load on the Server something has to take up the load... it cant just disappear. Even if it doesn't perceptively add load to the agent servers they might still be doing more work.
Comment