I am very happy to hear this, and even more so that BBWin now supports a centralized configuration.
Etienne writes:
I have also some general questions about bbwin :
Actually, svcs only works as a local agent because there is no option for it in hobbit-clients.cfg, it is very specific to Windows world. What do you prefer ?
- we keep svcs agent with local configuration. BBWin can handle centralized mode with agents which will run with local configuration
- we create a new option for the hobbit-clients.cfg which will be only used for Windows client
- May be you have other good ideas about that problem. At this time, it is only about svcs agent but for advanced features in the future, we will have other choices to make.
My preferred solution would be to have it in the hobbit-clients.cfg file, so there is a specific configuration statement for the Windows services check. There are a lot of Windows boxes being monitored with Hobbit, and I very much consider BBWin an essential companion to Hobbit. So the configuration should be as Hobbit-like as possible, even if it means implementing bits of code in Hobbit that are specific to BBWin.
With the centralized mode, you will see that you get graphs for each of your networks connections. However, for Windows, the names are too long compared to the unix name (eth0, eri0..). So, at this time, I use the mac address of the network card for the graph name. Would you prefer to use the name by making a very short name ? the ip address ? or is the mac address fine ?
There's really two sides to this issue. One thing is to identify the network interface from one poll to the next, so we put the right data into the RRD files - for this we can use any identifier that is guaranteed to be a)unique and b)permanent. Another thing is to present the interface name in some sensible way on the graph.
I don't really like using the MAC address, because it is difficult to relate to anything meaningful - I wouldn't know that my primary interface is 00:0E:A6:CE:D6:85, but I do know it's called "eth0". I've had the same problem for the SNMP data collection, so I have come up with an "rrd.meta" file that can map the ID of the network interface into a text that appears on the graph. The code isn't pretty, and it is static data that must be maintained by hand - if anyone has a better suggestion, please speak up.
There is a new agent 'who' used to report current connected users on the windows servers. There is no alert for it at this time. Which relevant type of alerts could be useful ? there is no alert options on Unix side. I was thinking of a rule like PROC but called USER with a minimum logged number and a maximum logged number for a username.
Would make sense, also as a check on the unix systems.
For 2008, I will do my best to create more releases.
Me too :-)
Regards, Henrik