Remotly Monitoring a Location reliably

[crs9]

I'm working on implenting zabbix in a service center environment and I want to see how others have handled this issue.

How do you get reliable uptime/ SLA numbers from the remote location if the WAN link fails?

Yes, I know a fail-over connection will help...lol. Just because the WAN link has failed, does not mean the end user at the location is having a loss of the internal services to them. Only way I see being able to fully hanlde this issue is for zabbix to collect the data on location and then forward it up to the "master/ central" server at my location. I have a server at each location already, so I can collect the data on it if needed. If I'm not making myself clear, please let me know. This was a problem with deploying nagios too in this environment, so I wouldn't say it's application specific. Thanks for any advice or feedback.

Chris

[Wolfgang]

I think there are several approaches:

a.) zabbix uses one local server on each location to collect data and then replicates to one central server.

b.) zabbix agents on different sites passes data to one central server

My understanding is that you want to go for a.)

As far as i am concerned zabbix does not provide a replication mechanism so far. So one option would be to replicate on mysql server level which would be possible.

Next problem then is that this would require one zabbix database on the central server for each local server because otherwise id's for items, hosts, users, triggers, data etc. will mess up.

This in turn would mean that there is no central gui on the central server that would cover all "satellite/local" server.

To work around this, the zabbix database scheme would have to be extended to support some sort of satellite/child/local server etc. as well as changes to gui daemon etc.

So in summary, i do not see an out-of-the-box solution so far except to go for b.) which is a standard zabbix setup and won't work if the wan link goes down.

[crs9]

Brainstorm

Yes I agree with your overview. That was basically the info I was coming up with, justing hoping someone else had a better answer. I just find it interesting that virtually no OSS has addressed the scenario a) that you outlined above.

Just total theorizing and bsing.. ..using the current design of Zabbix;

a) you would do everthing the same and set groups on all your devices based on location
ie: Main site, Remote 01, Remote 02
b) using the group flag, push the zabbix database and that noted group to the remote location's zabbix server (through MySQL replication)
c) the remote location's zabbix server would then ONLY push back the collected data on the devices at a user defined interval (through MySQL replication again)
d) even if the link was down, you can continue to collect all the proper data
e) admin of each location would ONLY be done at the central server, since it would be pushing out the location's device changes

Ideally there would be the central server and then a scaled back "remote" server just for collecting data and viewing that site's collected data.

Just thinking here about how it could be done, comments?

[Wolfgang]

@crs9
Maybe you want to spent a look on that draft/design:

http://www.zabbix.com/forum/showthread.php?t=2296