We have some test, dev and ci servers that we have setup as long running persistent spot instances mapped to specific domains using route53. This works great - we get the savings, we can allocate these without too much concern about cost but every now and then we loose the instance due to availability. When they come back - they come back with different IP addresses which breaks the route.
Is there a good way to have these instances automatically remap to the new IP address when they come back online (usually within a minute or two)?
Caution: I'm not convinced this approach is working after all. While I can confirm everything runs as I expected it to - the new routes didn't get setup correctly after these machines were assigned new spot instances. I'm not sure if this is because some service is not started by the time this script runs or if Amazon specifically prohibits this behavior. I'd be curious to hear what others have found.
--
N.B. The right answer here might well be using elastic IP addresses which as I understand allow you to have a single static IP address avoiding this issue altogether. I've not done the cost calculation on this but it might well be cheaper than the solution offered below.
What we ended up coming up with is a script that uses the AWS instance metadata and cli to make a route53 call upon reboot. This did NOT work on our old Ubuntu 14.04 instances but appears to on our newer Ubuntu 20.04 instances.
Here's how it works:
setupRoute53.shthat knows how to make a single call to route53.We're currently running this within the ubuntu user - the crontab looks like this:
And
setupRoute53.shlooks like this: