I recently was brought in to add capacity to an existing cluster of Wowza servers via Amazon EC2.
Due to project requirements, I needed to run the servers on a custom Ubuntu build, so I can't use the default images for Wowza provided. I also needed to use Elastic Block Store as opposed to instance storage, which the current Wowza AMI images use. Thus, I built my custom AMI on a stock Ubuntu server install provided by the Ubuntu team. I got a daily license, and installed it by default in my template AMI, and tested it to make sure things were working alright. I then went on to create about 20 images needed for the broadcast tonight.
After creating two instances, I noticed that something weird was happening: only one of them was showing up in the "loadbalancer?showServerXML" view! Then, I realized what was happening: both EC2 instances have the same "serverId" and thus after refreshing repeatedly, I'd see things alternate between my two servers. Basically, this means that a hundred servers can show up as 1 server to the load balancer, as they seem to have the same server id.
Is there a workaround for this? For now, my load-balancing for my EC2 instances seems to basically be a round-robin approach, where every time it's luck of the draw, rather than being properly balanced. It works, but not well. If two servers ping really close together, there's a good chance that one of them will never get any load.
Is this a known bug? Is there something I can do to mitigate this?