Why do I need to use two hostnames for origin tests?

The hostname of your origin is not intended to be accessible to end users. Most likely, requests to that hostname redirect to the CDN. If this is the case, origin tests need an alternative method. Here’s how it works:

  1. The SLA agent performs a DNS lookup on your origin server hostname, which you provide in the Hostname to Retrieve Origin IP Address field. This lookup returns your origin server’s IP address.
  2. The SLA agent forms a request for the SLA test object using that IP address.
  3. The SLA agent alters the HOST header value so it looks like a request to the hostname where users access your site.
  4. The SLA agent sends the modified HTTP request to your origin.

Because the headers no longer contain the protected origin hostname, the origin server allows the request and responds with the test object.

Why might the results show a possible outage or unsatisfactory performance when that’s not the case?

Some perfectly benign events can look like issues in your SLA reports. For example:

  • A problem might affect the measurement server itself and not the network it is testing.
  • If your origin is not reachable, the origin tests will report as failures. However, the corresponding CDN tests will not report as failures. Therefore, you might see a failure count for origin tests when the availability indicator is 100% because there was no SLA breach.
  • Performance tests require that the test object be cached for a specified period of time (see your SLA). If your Property Manager configuration does not allow for the necessary caching period, the performance test results might not be valid.

What should I do if the SLA tests indicate a potential issue?

If you think your site experienced either a potential availability outage or unsatisfactory performance, contact Support. Make sure you have this information available for your support team:

  • Your account name
  • The contract on which you detected an issue
  • The type of issue (availability or performance)
  • The name of the test where you detected an issue
  • The date(s)/time(s) of the data points that raised a concern for you; if there are many data points, a time range is sufficient

The support team will investigate the matter and respond to you.