Input on Distributed setup PLS!!!

Ok guys I have posted here about this beofre (sorry) but I need to make sure I have this all down before I restart this project again…

We have multiple data centers, and I would like to send checks to one central location… We have too much latency and iffy connections to send the payloads over directly from the clients… So… My plan is to have a Rabbitmq server at each location to (forward) the traffic on to the master location… At the master location I need a load balanced sensu cluster.

I currently have rabbitmq shovelers at the master location pulling the data from the remote locations and it works great. but that is to a single master.

Questions…

Is there a way to set up a cluster with rabbit, redis, and sensu while still using the current config (with the shovelers pulling from the remote locations)?

In a simple HA setup do you (on the client config) point the rabbitmq ip address to the HA proxy?

Are there any blogs about doing this in more detail than this one (http://failshell.io/sensu/high-availability-sensu/)?

All input is welcome and much appreciated!!

Also does anyone have a time frame for the release that will allow multiple sensu masters to report to one api???

  • Having a cluster of rabbitmq, redis, and sensu servers in once place with remote shovelers should work?
    You would have a central clustered Rabbitmq cluster in say, mirror mode, with the sensu cluster talking to it. Clients would be talking to the shovelers. Shovelers pump to the central cluster. The failshell article is a good (I think) example of how to build the central cluster part.

But I don’t think shovels are bi-directional, so it probably would only work with standalone checks? I don’t know about that. If you report it works, then apparently it does.

Have you looked at the Federation plugin?

  • I use round robin DNS for the clients, or you can point the clients to a vip and use something like keepalived. Pretty much any proven HA technology will work, it is tcp and the clients reconnect.

  • The failshell blog is the only one I know of, it is pretty complete. In the end it is about building reliable clusters of all the components, and then having each component talk to each other.

  • I don’t know about multiple masters reporting to one api. I thought it would be cool to have a dashboard that could consume multiple apis?

···

On Tue, Mar 25, 2014 at 4:14 PM, Joshua Zitting jzjoshzitting@gmail.com wrote:

Ok guys I have posted here about this beofre (sorry) but I need to make sure I have this all down before I restart this project again…

We have multiple data centers, and I would like to send checks to one central location… We have too much latency and iffy connections to send the payloads over directly from the clients… So… My plan is to have a Rabbitmq server at each location to (forward) the traffic on to the master location… At the master location I need a load balanced sensu cluster.

I currently have rabbitmq shovelers at the master location pulling the data from the remote locations and it works great. but that is to a single master.

Questions…

Is there a way to set up a cluster with rabbit, redis, and sensu while still using the current config (with the shovelers pulling from the remote locations)?

In a simple HA setup do you (on the client config) point the rabbitmq ip address to the HA proxy?

Are there any blogs about doing this in more detail than this one (http://failshell.io/sensu/high-availability-sensu/)?

All input is welcome and much appreciated!!

Also does anyone have a time frame for the release that will allow multiple sensu masters to report to one api???