Prevent healthcheck from being cached

[guilhermearaujo]

We run thumbor behind Cloudfront to cache the output images, without a direct, open route to the thumbor instances.

To monitor the application, we'd like to ping the health check route periodically, and since the only way for external tools to reach the server is going through Cloudfront, the health check route must no be cached.

This PR adds a cache control header with the no-cache policy to prevent Cloudfront (or whatever other service that may cache the result) from caching the response of the health check route.

[kkopachev]

Curious how you made your thumbor only respond to cloudfront. In order for CF to talk to your origin, it has to be publicly accessible.

[guilhermearaujo]

Using public load balancers and nginx, that re-route the requests to our applications in a private network.
We definitely could use the same nginx to inject this cache control header, what would be transparent when it gets back the CF, but I believe the application should be responsible for it.

But back to the PR, I can't understand why Travis is failing to compile pycurl 🤔
Even the master branch seems to be failing. And ideas?

[kkopachev]

I think load balancers should check for service health and remove unhealthy service from pool. Doing it on cloudfront level seems weird. https://docs.nginx.com/nginx/admin-guide/load-balancer/http-health-check/

[kkopachev]

In terms of travis failure. It seems that travis is randomly deciding to run a test on ubuntu xenial instead of default trusty. I made a PR to fix it #1207

[coveralls]

Coverage remained the same at 86.275% when pulling 150c4e9 on guilhermearaujo:healthcheck-cache into 37c3602 on thumbor:master.

[coveralls]

Coverage remained the same at 86.275% when pulling 150c4e9 on guilhermearaujo:healthcheck-cache into 37c3602 on thumbor:master.

[cristiandean]

Thank you @guilhermearaujo and @kkopachev!