Troubleshooting Zowe API Mediation Layer
Troubleshooting Zowe API Mediation Layer
As an API Mediation Layer user, you may encounter problems with how API ML functions. This article presents known API ML issues and their solutions.
To troubleshoot errors or warnings that can occur when configuring certificates, see the article Troubleshooting certificate configuration.
Install API ML without Certificate Setup​
For testing purposes, it is not necessary to set up certificates when configuring the API Mediation Layer. You can configure Zowe without certificate setup and run Zowe with zowe.verifyCertificates: DISABLED.
For production environments, certificates are required. Ensure that certificates for each of the following services are issued by the Certificate Authority (CA) and that all keyrings contain the public part of the certificate for the relevant CA:
- z/OSMF
- Zowe
- The service that is onboarded to Zowe
Enable API ML Debug Mode​
Use debug mode to activate the following functions:
- Display additional debug messages for API ML
- Enable changing log level for individual code components
- Gather atypical debug information
When on z/OS, API ML log messages are written to the STC job log.
We highly recommend that you enable debug mode only when you want to troubleshoot issues. Disable debug mode when you are not troubleshooting. Running in debug mode while operating API ML can adversely affect its performance and create large log files that consume a large volume of disk space.
Follow these steps:
-
Open the file
zowe.yaml. -
Each API ML component has its own
components.<component>.debugparameter. Set the value totruefor each component you want to debug. Note that there is no single setting that enables debug mode for all components at once.Example yaml to enable debug for all three core services:
components:gateway:debug: truediscovery:debug: truecatalog:debug: trueBy default, debug mode is disabled, and the
components.*.debugis set tofalse. -
Restart Zoweâ„¢.
You enabled debug mode for API ML core services (API Catalog, API Gateway and Discovery service).
-
(Optional) Reproduce a bug that causes issues and review debug messages. If you are unable to resolve the issue, create an issue here.
Change the Log Level of Individual Code Components​
You can change the log level of a particular code component of the API ML internal service at run time.
Follow these steps:
-
Enable API ML Debug Mode as described in Enable API ML Debug Mode. This activates the
/application/loggersendpoints in each API ML internal service (Gateway, Discovery service, and Catalog). -
List the available loggers of a service by issuing the GET request for the given service URL. Use the direct service URL when accessing a specific API ML component, or the Gateway-routed URL when accessing through the API Gateway:
GET scheme://hostname:port/application/loggers-
scheme
Specifies the API ML service scheme (http or https) -
hostname
Specifies the API ML service hostname -
port
Specifies the TCP port where API ML service listens on. The port is defined by the configuration parametercomponents.gateway.portfor the Gateway,components.discovery.portfor the Discovery service (by default, set to gateway port + 1), andcomponents.catalog.portfor the Catalog (by default, set to gateway port + 2).
DeprecatedThe following environment variables are deprecated. It is recommended to migrate to the corresponding
zowe.yamlparameters.Deprecated Environment Variable Replacement zowe.yamlParameterMFS_GW_PORTcomponents.gateway.portMFS_DS_PORTcomponents.discovery.portMFS_AC_PORTcomponents.catalog.portList loggers for API Catalog
To list the available loggers, issue a GET request for the given service URL in the following format:
GET [gateway-scheme]://[gateway-hostname]:[gateway-port]/apicatalog/api/v1/application/loggers -
One way to issue REST calls is to use the http command in the free HTTPie tool: https://httpie.org/.
Example:
# HTTPie (requires: pip install httpie)
http GET https://<gateway-hostname>:7554/application/loggers
# curl
curl -k https://<gateway-hostname>:7554/application/loggers | jq .
Output:
{"levels":["OFF","ERROR","WARN","INFO","DEBUG","TRACE"],
"loggers":{
"ROOT":{"configuredLevel":"INFO","effectiveLevel":"INFO"},
"com":{"configuredLevel":null,"effectiveLevel":"INFO"},
"com.ca":{"configuredLevel":null,"effectiveLevel":"INFO"},
...
}
}
Filter logger output with grep to narrow down specific packages, for example:
http GET https://<gateway-hostname>:7554/application/loggers | grep -i "zowe"
-
Alternatively, extract the configuration of a specific logger using the extended GET request:
GET scheme://hostname:port/application/loggers/{name}-
{name}
Specifies the logger name
-
-
Change the log level of the given component of the API ML internal service. Use the POST request for the given service URL:
POST scheme://hostname:port/application/loggers/{name}The POST request requires a new log level parameter value that is provided in the request body:
{"configuredLevel": "level"}-
level
Specifies the new log level: OFF, ERROR, WARN, INFO, DEBUG, TRACE
Example:
# HTTPie (requires: pip install httpie)http POST https://hostname:port/application/loggers/org.zowe.apiml.enable.model configuredLevel=WARN# curlcurl -k -X POST https://hostname:port/application/loggers/org.zowe.apiml.enable.model \-H "Content-Type: application/json" \-d '{"configuredLevel": "WARN"}'HTTPie vs curl comparison:
Tool Installation Command syntax Use case HTTPie pip install httpiehttp POST <url> key=valueHuman-friendly, auto-formatted output curl Pre-installed on most systems curl -X POST <url> -H ... -d '...'Universal, scriptable, no extra dependencies Hierarchical logging:
Loggers in API ML follow a hierarchical naming convention. Setting the log level on a parent logger (for example,
org.zowe.apiml) affects all child loggers (org.zowe.apiml.gateway,org.zowe.apiml.security, etc.) unless a child has an explicitly configured level.Common loggers:
Logger Description org.zowe.apimlAll API ML components (Gateway, Discovery, Catalog) org.zowe.apiml.gatewayGateway-specific components org.zowe.apiml.securitySecurity and authentication components com.netflix.eurekaEureka service discovery and registration org.springframeworkSpring framework internals org.apache.httpHTTP client and wire-level requests reactor.nettyReactive Netty I/O operations Runtime-only persistenceLog level changes made via the
/application/loggersendpoint apply only for the current session. They are not persisted across restarts. To make permanent changes, setcomponents.<component>.debug: trueor configure logging in the component'sapplication.yml. -
Gather atypical debug information​
Use the following configuration to set either verbose internal logging for key system packages, or enable detailed SSL/TLS tracing to analyze encrypted traffic layers.
- debug
This boolean property activates the Springdebugprofile for an API ML component and enables the/application/loggersendpoint at runtime. Enabling this property sets verboseDEBUGorTRACElog levels for key packages, includingorg.zowe.apiml,org.springframework,org.apache.http, andreactor.netty.
Set debug to true under the relevant component in zowe.yaml:
components:
gateway:
debug: true
To enable debug logging on the API ML single-service, set debug to true under apiml:
components:
apiml:
debug: true
For more information, see Spring Boot Profiles in the Spring documentation.
When debug: true is set, you may not see the debug output immediately in the STC job log on z/OS or in the container logs. Check the component's application log file (for example, $WORKSPACE_DIR/.logs/gateway/) for the detailed debug messages. On z/OS, the debug output is written to the job log of the started task.
- sslDebug
This property enables SSL/TLS debug logging and can assist with determining what is happening at the SSL layer. This property maps directly to the-Djavax.net.debugJava system property.
Set sslDebug under the relevant component in zowe.yaml. The following example enables full SSL debug logging on the Gateway:
components:
gateway:
sslDebug: "ssl"
To enable SSL debugging on the API ML single-service deployment (which bundles Gateway, Discovery, API Catalog, and ZAAS), set sslDebug under apiml:
components:
apiml:
sslDebug: "ssl"
The value ssl turns on all SSL debugging. For finer-grained tracing, the property accepts the same sub-options as the standard javax.net.debug Java property, for example ssl:record,handshake or all.
For more information, see the article Debugging Utilities in the IBM documentation.
The sslDebug property can also be enabled for other API ML components.
Services that are not running appear to be running​
Services that are not running appear to be running. The following message is displayed in the Discovery service:
EMERGENCY! EUREKA MAY BE INCORRECTLY CLAIMING INSTANCES ARE UP WHEN THEY'RE NOT. RENEWALS ARE LESSER THAN THRESHOLD AND HENCE THE INSTANCES ARE NOT BEING EXPIRED JUST TO BE SAFE.
Cause:
This message is expected behavior of the Discovery service. If a service is incorrectly terminated without properly unregistering from Eureka, the service initially enters eviction status for a brief timeframe before the service is deregistered. Failure to properly terminate occurs when a service fails to respond to three consecutive heartbeat renewals. After the three heartbeat renewals are returned without a response, the Eureka Discovery service keeps the service in eviction status for one additional minute. If the service does not respond within this minute, the Eureka service unregisters this unresponsive service. When more than 15 percent of currently registered services are in eviction status, self preservation mode is enabled. In self preservation mode, no services in eviction status are deregistered. As a result, these services continue to appear to be running even though they are not running.
Solution:
Use one of the following options to exit self preservation mode:
-
Restart the services that appear to be running
Relaunch the services that appear to be registered. After the message disappears, close each of the services one at a time. Allow for a 3-minute period between closing each service. The procedure for restarting services that are not part of Zowe is specific to the services and is documented in the service documentation. -
Restart the Discovery service
Manually restart the Discovery service. The new instance will not be in self preservation mode. In a few minutes, the running services re-register.Note:
The Discovery service can be stopped with the following command:
F <instance-job-name>,APPL=STOP(<component_name>)The Discovery service can be started again with the following command:
F <instance-job-name>,APPL=START(<component_name>)Example:
F ZWESLSTC,APPL=STOP(discovery-service)F ZWESLSTC,APPL=START(discovery-service) -
Adjust the threshold of services in eviction status
Change the frequency of the Discovery service from entering self preservation mode by adjusting the threshold of services in eviction status.Note: The default threshold is .85. This results in the Discovery service entering self preservation mode when 15 percent of currently registered services are in eviction status.
Example:
eureka.renewalPercentThreshold=0.3This threshold limit causes the Discovery service to enter self preservation mode when less than 30 percent of services are not responding.
Debug and Fix Common Problems with SSL/TLS Setup​
Review tips described in the blog post Troubleshooting SSL/TLS setup with Zowe Certificate Analyzer to find out how you can use the Zowe Certificate Analyzer to address the following common issues with SSL/TLS setup:
- How to verify if the API ML server certificate is trusted by your service
- How to get a CA certificate in the correct format
- How to perform a TLS handshake with debug logs
- How to debug remote services
- How to enable mutual authentication using a client certificate
- How to add a trusted certificate to a SAF Key ring
SDSF Job search fails​
Search for jobs using SDSF failed for prefix and owner : exc.sdsf_invocation_failed 8 (Issue does not impact ZD&T boxes)
Solution:
You must be authorized to use SDSF with REXX on your z/OS system. For authorization, activate the SDSF RACF class and add the following 3 profiles to your system:
GROUP.ISFSORIGGROUP.ISFSPROG.SDSFISF.CONNECT.
Users must belong to a group that has READ access to these profiles.
This is quite a complex area and you should ask your systems programmer for advice. On most systems, the GROUP.* profiles are not required and it is sufficient to have the following ISF profile defined:
class profile SDSF ISF.CONNECT.** (G)