eclipsefdn / cerberus Goto Github PK

Marketplace.eclipse.org was not reachable by our users this morning. However, Eclipse Status did not notice any downtime.

We believe this is caused by our Nginx proxy server returning a valid cached response if the service is down.

A possible solution would be to query a different URL randomly each time: https://marketplace.eclipse.org?status=:unique_random_string

Nginx will try to fetch the resource from the webserver since it's a new URL even we are still really fetching the homepage. I don't believe this issue is limited to Marketplace or Drupal sites.

Another solution would be to use specific request headers that we already have in place to bypass/revoke the Nginx cache for a particular page/resource.