Multiple active hosts on one physical machine?
1st, Thanks for zabbix team provide such a good monitor tools.
I deployed a agent on a physical machine, but have multiple server on it. So I create multiple hosts on the server side.
The problem is, when I need to monitor several log files under different hosts, but zabbix only allows one host do the active check.
I examine agent code, and find the inner mechanism is:
1. At initialization, zabbix agent fork a process for active checks.
2. The process get host name from zabbix_agentd.conf, from Hostname=xxx
3. The process request the server cyclically, get all active checks from server's database by hostname of xxx.
4. The process run all the active checks, and send the result back to server, with the hostname of xxx.
So I think, there a 3 ways to solve my problem
1. Put all active checks under one host of the physical machine. This solution need not to change the code, but when trigger is on, on the server side we can't easily find which host is actually in problem.
2. Start multiple agents on a physical machine with different config file. Each binds to one listening port and set one hostname. Again no need to change the code. But if I have 20 host on one machine, I need to start over 100 zabbix_agentd process for monitor.
3. Change The code, which read config file like this:
hostname=xxx,yyy,zzz
and start 3 active check process corresponding to each hosts. and do the same steps as before. This maybe the best solution.
I searched the forum and didn't find anyone has the same problem with me . If there is any discussion or better solution, please tell me. Or do zabbix team think the 3rd way is a good feature to be implemented in future?
Many thanks.
I deployed a agent on a physical machine, but have multiple server on it. So I create multiple hosts on the server side.
The problem is, when I need to monitor several log files under different hosts, but zabbix only allows one host do the active check.
I examine agent code, and find the inner mechanism is:
1. At initialization, zabbix agent fork a process for active checks.
2. The process get host name from zabbix_agentd.conf, from Hostname=xxx
3. The process request the server cyclically, get all active checks from server's database by hostname of xxx.
4. The process run all the active checks, and send the result back to server, with the hostname of xxx.
So I think, there a 3 ways to solve my problem
1. Put all active checks under one host of the physical machine. This solution need not to change the code, but when trigger is on, on the server side we can't easily find which host is actually in problem.
2. Start multiple agents on a physical machine with different config file. Each binds to one listening port and set one hostname. Again no need to change the code. But if I have 20 host on one machine, I need to start over 100 zabbix_agentd process for monitor.
3. Change The code, which read config file like this:
hostname=xxx,yyy,zzz
and start 3 active check process corresponding to each hosts. and do the same steps as before. This maybe the best solution.
I searched the forum and didn't find anyone has the same problem with me . If there is any discussion or better solution, please tell me. Or do zabbix team think the 3rd way is a good feature to be implemented in future?
Many thanks.
Comment