The Repose filter chain is the work horse of the system, and is responsible for nearly all enrichment that Repose provides. It’s based on and uses the same contract as the JEE Filter Chain. As the name implies it is a chain of filters that are followed in order. A request goes through them from the first, one after another, until the last is reached when the request which has possibly been mutated along the way is sent on to the origin service. When the response comes back from the origin service it works it’s way back up the chain in the opposite order.

The Repose filter chain provides a number of enhancements over the standard filter chain contract that gives a lot of power and extra value.

Dynamic Filtering

Typically a filter chain is static and unchanging. Your request will pass through every single filter everytime regardless of whether you want it to or not. The Repose filter chain allows you to determine if a filter should be run while the request is processing.

URI Regex

This method of determining if a filter should be run is DEPRECATED and will be removed in Repose 10.

You can decide whether or not a filter should run based on the path of it’s URI using a regular expression. Here’s a simple example:

system-model.cfg.xml (partial)
...
<filters>
    <filter name="keystone-v2" uri-regex=".*/foo"/> (1)
    <filter name="api-validator"/> (2)
</filters>
...
1 This filter will only be run when the request is to a resource ending in /foo.
2 This filter will be run everytime.

It’s worth noting that if anything changes the request uri while processing the filter chain that will potentially impact whether a filter is run. Or in other words the filters to be run is not determined up front, but instead evaluated when the request hits a given filter in the chain.

Filter Activation Determination

You can configure whether or not a filter should run based on each of the following HTTP Request criterion:

This is further customizable through the use of the following boolean logic operators:

  • And

  • Or

  • Not

The filter chain is dynamic. It is not determined at the start of the request, rather before each filter is executed the request is compared against the configured criteria. This means that anything that modifies the request may effect whether a later filter will process or not. For example, the Body Extractor to Header Filter could add a header based on the request body which is used to determine if a later filter executes.

Refer to System Model Filter Activation Determination for further details about configuring this capability.

Intrafilter Logging

Sometimes you won’t get the output you expect from the filter chain whether that’s to your origin service or your customer that originated the request. Intrafilter logging will let you get a look into what effect each filter is having on both the request and response. All you have to do is change this line in your logging configuration, from:

log4j2.xml (partial)
...
<Logger name="intrafilter-logging" level="info"/>
...

to:

log4j2.xml (partial)
...
<Logger name="intrafilter-logging" level="trace"/>
...

This will convert both the request and response into a simple JSON object that will be written out into the logs. And it will do this before each filter and the origin service allowing you to see how each filter mutated them.

This can be expensive in both time and disk space and is intended only as a debugging tool, not for full time use.

For more troubleshooting help see our guide.

Metrics

Repose automatically records metrics on how long each request spends within a filter. If you wish to see these numbers or exfiltrate them to another tool look into the Metrics Service. They are listed under org.openrepose.core.FilterProcessingTime.Delay with an entry for each filter by name.

If you are curious about the numbers for a specific request you can add the X-Trace-Request header to the request. See Time Spent in Each Filter for more details.

We also start an OpenTracing span for each filter that is ran. If you’d like to get this data look into the OpenTracing Service.

Bypass URI

Sometimes you’d like a request to entirely bypass the filter chain entirely. The most common use case for this is if you have some sort of monitoring setup to check the responsiveness of your origin service. We support this with a simple regular expression specified at the top of the filter chain, like so:

system-model.cfg.xml (partial)
...
<filters bypass-uri-regex="/healthcheck">  (1)
    <filter name="keystone-v2"/>
    <filter name="api-validator"/>
</filters>
...
1 Any call to /healthcheck will skip the entire filter chain and go straight to the origin service.