ctennis's questions

I have an haproxy frontend that is basically redirecting to multiple backends. The logic to decide which backend to go to is outside of haproxy's control at the moment. However, what I would like to do (if haproxy will support it) is something like this:

frontend main localhost:4443
  default_backend be

backend be
  server be1 10.10.10.10:443 

And let my listening server at 10.10.10.10 issue a 302 redirect to forward the request on to the correct server.

However, I don't want the person accessing the front end to see this. I want the backend issuing the 302 back to haproxy, and haproxy hitting the new URL and forwarding that back to the frontend. Caching this would also be nice.

Is this feasible with haproxy? I can't tell from the configuration if this would work or not.

I am running a number of java processes on a single Linux machine. From a memory and computing standpoint, everything is fine when things are static.

However, periodically we use a configuration management package up upgrade the jar or war files, and restart the java process.

The problem is, that is restarts them all relatively quickly, and so we get 10 or so java VMs restarting all at the same time (we use daemontools for the service stops/starts), which wreaks havoc on the machine, in terms of OOMs or just really slow. This is because it's spawning the JVM 10x at the same time.

Other than trying to stagger the startups, is there a smarter way of handling this? Maybe a sysctl tuning performance parameters, or a JVM parameter?

I have a multi-homed network setup. I'm using a Cisco 7200 to do BGP advertising to both provider A and B. For each provider, I have a separate /24 that I'm using. I am AS prepending to influence the block A to provider A, and block B to provider B. This works fine. If either provider goes down, failover works great.

However, I also want to influence OUTBOUND traffic to flow to each provider in the same way. That is, as traffic comes in my router, I want to associate its outflow to the provider based on the source IP address. As is, each provider advertises to me just a default route.

I believe I can use routing policy to achieve this, but I don't want to hardcode the IP addresses in the policy since they could change dynamically.

So, what are my options to influence the outbound routing?

I'm curious if any of you have used any of the non-amazon-provided EC2 kernels with success? Canonical seems to provide newer kernels (2.6.28, eg), but there are some mixed reviews about their successful usage on the EC2 hardware.

I'm mainly interested in newer versions, not necessarily 2.6.18 stuff from other vendors (unless there is a smart reason to be using it).