The average number of transactions, meaning API proxy requests, per second. Note that if you have a relatively low number of transactions over the time period, the average number of transactions per second could appear to be zero in UI custom reports if the number is smaller than two decimal places.

API syntax:tps

The number of successful API requests that use the Response Cache instead of the response from the target service.

API syntax:sum(cache_hit)

Returns the number of elements in L1 (in-memory) cache per transaction over a given time period. For example, if you choosemax for the period of a day, and within that day the highest number of elements in cache is 12 for a specific transaction, then the count will be 12. Foravg, if there are three transactions in the time period you're querying, and their cache counts are 5, 6, and 7, the average is 6. L1 cache is in-memory cache as opposed to L2 database cache, as described inCache internals.

API syntax:avg(ax_cache_l1_count)

The total number of policy errors over the specified time period.

Policy errors usually occur by design. For example, the Verify API Key policy throws an error when an invalid API key is passed in the request, and a Spike Arrest policy throws an error if the number of API calls exceeds the limit defined in the policy. So this metric is useful for finding potential trouble spots in your APIs. For example, policy_error metrics, grouped by the developer_app dimension, might help you discover that an API key or OAuth token has expired for a given app; or you might find that a specific API proxy is throwing a lot of Spike Arrest errors, leading you to discover that the proxy's spike arrest limit doesn't account for an increase in holiday traffic.

A policy error is logged in analytics only if the error results in API proxy failure. For example, if a policy'scontinueOnError attribute is set totrue, the API proxy continues processing a request even if the policy fails. In that case, a policy error is not logged in analytics.

The Policy Name on Error (ax_execution_fault_policy_name) dimension is useful for grouping policy errors by policy name.

A target failure (such as a 404 or 503) does not count as a policy failure. Those count as API proxy failures (is_error).

API syntax:sum(policy_error)

The total number of times API proxies failed over the specified time period. Proxy failure can occur when a policy fails or when there's a runtime failure, such as a 404 or 503 from the target service.

The Proxy (apiproxy) dimension is useful for grouping API proxy failures by proxy.

API syntax:sum(is_error)

The amount of time (average, minimum, or maximum), inmilliseconds, that it takes Edge to process incoming requests. The time starts when the request reaches Edge and ends when Edge forwards the request to the target service.

Using different dimensions, you can examine request processing latencies by API proxy, developer app, region, and so on.

API syntax:max(request_processing_latency)

The size of the request payload received by Edge, inbytes.

API syntax:avg(request_size)

The total number of times a Response Cache policy was executed over the given time period.

Since the Response Cache policy is attached in two places in an API proxy (once in the request and once in the response), it usually executes twice in an API call. A cache 'get' and a cache 'put' each count as one execution.

However, response cache execution is 0 if the<SkipCacheLookup> element in the policy evaluates to true (in the request), and 0 if the<SkipCachePopulation> element in the policy evaluates to true (in the response).

In theTrace tool, you can click the Response Cache icon in an executed API call and view theresponsecache.executedflow variable to see whether or not there was a cache execution (a value of 1).

API syntax:sum(ax_cache_executed)

The amount of time (average, minimum, or maximum), inmilliseconds, that it takes Edge to process API responses. The time starts when the API proxy receives the target service response and ends when Apigee forwards the response to the original caller.

Using different dimensions, you can examine response processing latencies by API proxy, region, and so on.

API syntax:min(response_processing_latency)

The size of the response payload returned to the client, inbytes.

API syntax:max(response_size)

The total number of 5xx responses from the target service. These are target service errors not caused by Apigee.

API syntax:sum(target_error)

The amount of time (sum, average, minimum, or maximum), inmilliseconds, for the target server to respond to a call. This metric tells you how target servers are performing. The time starts when Edge forwards a request to the target service and ends when Edge receives the response.

Note that if an API call returns a response from cache (using the Response Cache policy, for example), the call will never reach the target service, and no target response time metrics are logged.

API syntax:avg(target_response_time)

The amount of time (sum, average, minimum, or maximum), inmilliseconds, from when Edge receives a request from a client to when Edge sends the response back to the client. The time includes network overhead (such as the time it takes load balancers and routers to do their work), request processing latency, response processing latency, and target response time (if the response is served from the target service instead of cache).

Using different dimensions, you can examine processing latencies by API proxy, developer app, region, and so on.

API syntax:avg(total_response_time)

The total number of API calls processed by Edge in the specified time period.

Use dimensions to group traffic counts in ways that are most meaningful to you.

API syntax:sum(message_count)

The name of the API product containing the API proxies being called. In order to get this dimension, developer apps making the calls must be associated with one or more API products that contain the API proxies, and the proxies being called must check for an API key or OAuth token sent with the API call. The key or token is associated with an API product. For more information, seeFirst things first: How to generate complete analytics data.

If the above criteria aren't met, you'll see the value "(not set)". See alsoWhat does an analytics entity value "(not set)" mean?.

The key containing the Response Cache value that was accessed. For more information on how the key is constructed for response cache, seeResponse Cache policy.

In theTrace tool, when you select a Response Cache policy that read from or wrote to cache, you can see this value in theresponsecache.cachekeyflow variable.

The name of the cache containing the keys/values used by the Response Cache policy, prefixed withorgName__envName__. For example, if the org is "foo," the environment is "test," and the cache name is "myCache," the ax_cache_name is foo__test__myCache.

In theTrace tool, when you select a Response Cache policy, you can see this value in theresponsecache.cachenameflow variable.

The cache level ("L1" in-memory or "L2" database) from which the Response Cache was retrieved. This dimension also shows "CACHE_MISS" when the response was delivered from the target instead of cache (and response cache was refreshed with the target response); or when a cache key in the request is invalid. Cache keys are limited to 2 KB in size.

In theTrace tool, when you select the Response Cache policy, you can see this value in theresponsecache.cachesourceflow variable.

For more information on cache levels, seeCache internals.

The consumer key (API key) of the developer app making the API calls, whether passed in the request as API keys or included in OAuth tokens.

In order to get this dimension, proxies receiving calls must be configured to check for a valid API key or OAuth token. Developer apps get API keys, which can be used to generate OAuth tokens, when the apps are registered in Edge. For more information, seeFirst things first: How to generate complete analytics data.

If the above criteria aren't met, you'll see the value "(not set)". See alsoWhat does an analytics entity value "(not set)" mean?.

The Edge-registered developer app making API calls.

In order to get this dimension, apps must be associated with one or more API products that contain the API proxies being called, and the proxies must check for an API key or OAuth token sent with the API call. The key or token identifies the developer app. For more information, seeFirst things first: How to generate complete analytics data.

If the above criteria aren't met, you'll see the value "(not set)". See alsoWhat does an analytics entity value "(not set)" mean?.

The email of the Edge-registered developers whose app made the API calls.

In order to get this dimension, developers must have apps associated with one or more API products that contain the API proxies being called, and the proxies must check for an API key or OAuth token sent with the API call. The key or token identifies the developer app. For more information, seeFirst things first: How to generate complete analytics data.

If the above criteria aren't met, you'll see the value "(not set)". See alsoWhat does an analytics entity value "(not set)" mean?.

The unique Edge-generated developer ID in the form oforg_name@@@unique_id.

In order to get this dimension, developers must have apps associated with one or more API products containing the API proxies being called, and the proxies must check for an API key or OAuth token sent with the API calls. The key or token identifies the developer. For more information, seeFirst things first: How to generate complete analytics data.

If the above criteria aren't met, you'll see the value "(not set)". See alsoWhat does an analytics entity value "(not set)" mean?.

Thefault code of the error. For example:messaging.adaptors.http.flow.GatewayTimeout

The namedflow in an API proxy that raised an error. For example, "PreFlow," "PostFlow," or the name of a conditional flow you created.

Note that the full name to use in the management API is ax_execution_fault_flow_name, without a line break.

Where no errors occurred, you'll see the value "(not set)".

The name of the API proxy flow states that raised errors, such as "PROXY_REQ_FLOW" or "TARGET_RESP_FLOW."

Note that the full name to use in the management API is ax_execution_fault_flow_state, without a line break.

The name of the policy that threw an error and caused the API call to fail.

Note that the full name to use in the management API is ax_execution_fault_policy_name, without a line break.

If a policy throws an error but the policy root attributecontinueOnError is set totrue, the API proxy flow continues without failure, and the policy failure isn't counted in this dimension.

The BasePath configured on the API proxy ProxyEndpoint. Base path does not include the domain and port portion of the API proxy URL. For example, if an API proxy's base URL is https://apigeedocs-test.apigee.net/releasenotes/, the base path is /releasenotes.

The value is also stored in theproxy.basepathflow variable.

The resource path added to the API proxy base path. For example, if an API proxy's base URL ishttps://apigeedocs-test.apigee.net/hello/, and a call is made tohttps://apigeedocs-test.apigee.net/hello/json, the pathsuffix is/json.

If no pathsuffix is used, the value is empty.

The value is also stored in theproxy.pathsuffixflow variable.

Contains the originating client IP address. The value of theax_resolved_client_ip dimension is calculated from the values in theax_true_client_ip and thex_forwarded_for_ip dimensions.

Note that when using routing products such as Akamai to capture the true IP addresses of clients, the client IP is passed to Edge in the HTTP headerTrue-Client-IP, which is then used to set theax_true_client_ip dimension.

The value of theax_resolved_client_ip dimension is calculated as follows:

1. Ifax_true_client_ip is not null and does not contain a local IP address, then setax_resolved_client_ip toax_true_client_ip.
2. Else, setax_resolved_client_ip to the first non-local IP address inx_forwarded_for_ip.
3. If bothax_true_client_ip andx_forwarded_for_ip contain only local IP addresses, then setax_resolved_client_ip to the first local IP address inx_forwarded_for_ip.
4. If bothax_true_client_ip andx_forwarded_for_ip are null, setax_resolved_client_ip to(not set).
5. Ifax_true_client_ip is a local IP address andx_forwarded_for_ip is null, setax_resolved_client_ip to(not set).

The operating system version of the device making the call.

It's useful to use this as a second "drill-down" dimension with OS Family (ax_ua_os_family) to see the versions of the operating systems.

The IP address of the calling client, stored in theproxy.client.ipflow variable. This is often theX-Forwarded-For address of the inbound call, which is the IP address Edge received from the last external TCP handshake. This could be the calling client or a load balancer. When there are multiple IPs in theX-Forwarded-For header, this is the last IP listed.

When using routing products such as Akamai to capture the true IP addresses of clients, the client IPs are passed to Edge in the HTTP headerTrue-Client-IP. This dimension captures those true client IPs from that header.

To determine the original client IP Address, accessed through theax_resolved_client_ip dimension, Edge uses theax_true_client_ip and thex_forwarded_for_ip dimensions.

The resource path (not including the domain) to the target service, excluding query parameters.

For example, the Apigee sample targethttp://mocktarget.apigee.net includes several resources, including/user, which returns a greeting. Regardless of how your API proxy callshttp://mocktarget.apigee.net/user, the request_path is/user.

The resource path (not including the domain) to the target service, including query parameters.

For example, the Apigee sample targethttp://mocktarget.apigee.net includes several resources, including/user?user={name} resource and query parameter to return a custom greeting to the name provided. Regardless of how your API proxy callshttp://mocktarget.apigee.net/user?user=Dude, the request_uri is/user?user=Dude.

The name of the user agent, or software agent, used to make the API call. Examples:

• A Pixel XL making a call through Chrome:Mozilla/5.0 (Linux; Android 7.1.2; Pixel XL Build/NHG47N) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/59.0.3071.92 Mobile Safari/537.36
• An iPad making a call through Chrome:Mozilla/5.0 (iPad; CPU OS 10_2 like Mac OS X) AppleWebKit/602.1.50 (KHTML, like Gecko) CriOS/54.0.2840.91 Mobile/14C92 Safari/602.1
• cURL from a terminal:curl/7.51.0

The version of the useragent.

It's useful to use this as a second "drill-down" dimension with User Agent Family (ax_ua_agent_family) to get the version of the agent family.

The resource path (not including the domain) to the target service, excluding query parameters, that is defined in the proxy's<TargetEndpoint>.

For example, say an API proxy calls the following target:

<TargetEndpoint name="default">...<HTTPTargetConnection>  <URL>http://mocktarget.apigee.net/user?user=Dude</URL></HTTPTargetConnection>

In this example, the target_basepath is/user.

If the target were this:

<TargetEndpoint name="default">...<HTTPTargetConnection>  <URL>http://mocktarget.apigee.net</URL></HTTPTargetConnection>

the target_basepath would be null.

In theTrace tool, when you select the AX icon at the end of the flow diagram, thetarget.basepathflow variable maps to the target_basepath dimension.

The HTTP response status code returned by the target service to the API proxy, such as 200, 404, 503, and so on.

A value of "null" means the request never reached the target service. This occurs when the response is served by the Response Cache policy or when there's a failure in request processing.

This is different than theResponse Status Code (response_status_code) dimension.

The full URL of the target service defined in an API proxy's TargetEndpoint.

<TargetEndpoint name="default">...<HTTPTargetConnection>  <URL>http://mocktarget.apigee.net/user?user=Dude</URL></HTTPTargetConnection>

In this example, the target_url ishttp://mocktarget.apigee.net/user?user=Dude.

Note that the URL can also be overridden during API proxy processing with thetarget.urlflow variable.

Inproxy chaining and when usingscript targets (Node.js), the target_url in the calling proxy is null.

The list of IP addresses in theX-Forwarded-For header.

To determine the original client IP Address, accessed through theax_resolved_client_ip dimension, Edge uses theax_true_client_ip and thex_forwarded_for_ip dimensions.

Based on a 24-hour clock, the 2-digit hour in which API calls were made. For example, API calls made in the hour between 10pm and 11pm, the ax_hour_of_day would be 22.

The time value is in UTC.