One of the things we really make use of in our current bb setup is the failover capability, which has been quite useful on more than one occasion.
We have 2 geographically separate data centers which are connected via a fat pipe, and we do redundant monitoring of each data centers via bb servers from each data center. The bb servers in data center 1 normally does all the alerting, but in case the bb servers in data center 2 can't reach the bb servers in data center 1 (for instance in case of wan outage) the bb servers in data center 2 fail over and take on all the alerting duties. If and when the bb servers in data center 1 become reachable again, the bb servers in data center 1 revert to the normal non-alerting state.
I looked through the hobbit configs but don't see that failover functionality. Is that something envisioned for a future release, or is there simply a different approach to that problem in hobbit as compared to bb?
On Sat, Jun 10, 2006 at 08:41:30PM -0700, J Sloan wrote:
One of the things we really make use of in our current bb setup is the failover capability, which has been quite useful on more than one occasion. [snip] I looked through the hobbit configs but don't see that failover functionality. Is that something envisioned for a future release, or is there simply a different approach to that problem in hobbit as compared to bb?
In the Hobbit "contrib" directory, you will find a detailed description of how you can setup a cluster of Hobbit servers for an HA scenario.
I do have some ideas about other ways of handling this, but it is fairly low on my priority list since the cluster mechanism works and can be setup without any changes to Hobbit itself.
Regards, Henrik
Henrik Stoerner wrote:
On Sat, Jun 10, 2006 at 08:41:30PM -0700, J Sloan wrote:
One of the things we really make use of in our current bb setup is the failover capability, which has been quite useful on more than one occasion. [snip] I looked through the hobbit configs but don't see that failover functionality. Is that something envisioned for a future release, or is there simply a different approach to that problem in hobbit as compared to bb?
In the Hobbit "contrib" directory, you will find a detailed description of how you can setup a cluster of Hobbit servers for an HA scenario.
This is a nice document, but unfortunately the method described won't work for us as we have separate data centers and can't have a virtual server straddling the different networks with a single IP.
Also, we need for both servers to be running all the time, doing the tests, updating the web pages etc, but only one side is to send notifications. The notifications are precisely the part that needs to fail over, and the way BB does it is perfect for making this happen.
Perhaps after graduation this fall I can devote some time to working on that, so that we could replace the aging bb with hobbit.
Joe
participants (2)
-
henrik＠hswn.dk
-
joe＠tmsusa.com