Known limitations
This page describes the known limitations for API Connect 10.0.7.0 and later.
- Issues transforming database encryption during upgrade operation to v10.0.8.2
- IBM has identified and addressed an issue that may occur during upgrades to API Connect version
10.0.8.2, affecting database encryption transformation.
To resolve this issue, we have released v10.0.8.2-ifix1 as a full installation image.
- New deployments: Perform a fresh installation using v10.0.8.2-ifix1 to avoid encountering this issue.
- Existing deployments: If you have already upgraded to v10.0.8.2 and are experiencing this problem, upgrade directly to v10.0.8.2-ifix1 ( a complete replacement, not an incremental patch). If you have not yet upgraded to 10.0.8.2 then upgrade directly to v10.0.8.2-ifix1.
Before upgrading to APIC v10.0.8.2-ifix1, you must run the
apicops preupgrade
command to check for:- Recent encryption-related issues
- Invalid gateway-extensions
For detailed upgrade or installation steps, refer to the IBM documentation.
- WebSocket support
-
API Connect provides basic WebSocket functionality through the
websocket-upgrade
policy. However, there are important limitations to consider when designing APIs that use WebSocket communication.- General limitations
- The maximum reliable message payload is less than 1 KB.
- Sending several messages in rapid succession may exceed cumulative limits. For example, three consecutive 500-byte messages may fail.
- Messages up to 100 bytes can be reliably transmitted if sent with at least a 10 millisecond delay between them.
- Send data at intervals greater than 10 milliseconds to prevent dropped connections or gateway issues.
- WebSocket compression and streaming modes are not supported.
- Error handling limitations
- When a server closes a connection using status code 1000, 1006, or 1008, the client receives code 1006, regardless of the actual cause.
- If the server crashes, the client does not receive a close event or error message from the DataPower Gateway.
- Errors can be caught:
- In the main assembly before the
websocket-upgrade
policy is executed. - In sub assemblies during message processing after the upgrade that originate from a subassembly action.
- In the main assembly before the
- Errors during WebSocket connections or after disconnection cannot be detected once the upgrade is complete.
- Opcode and data type support
- Only text frames are supported.
- Binary frames are not supported.
- Policy restrictions after upgrade
- After the WebSocket upgrade, the following actions and properties are not supported in the sub assemblies:
- client security
- generate JWT
- user security
- validate JWT
- client identification
- activity log
- HTML page
- WebSocket upgrade
- Parse:
- use content type
- warn on empty input
- No actions are allowed in the main assembly after the upgrade.
- Only sub assemblies that use supported properties can process messages after the upgrade.
- After the WebSocket upgrade, the following actions and properties are not supported in the sub assemblies:
- General limitations
- Warning message after upgrading from version 10.0.5.x to version 10.0.8.x
-
After upgrading from API Connect v10.0.5.x to v10.0.8.x, users encounter the following warning message in the analytics component:
Warning: The transform of API event data into long term summary data has encountered problems. Contact your system administrator to investigate further. Long term summary data in reports may not be complete.
This warning is displayed because API event data processed by v10.0.5.x cannot be transformed into long-term summary data that uses v10.0.8.x.
The warning does not affect analytics functions or the accuracy of newly processed data in version 10.0.8.x. To resolve the warning automatically, ensure that the data retention policy is properly configured. As older data is purged, the warning clears. To update the retention settings, see Retention and rollover.
- Validating from within a ruleset shows all rules
- In the governance service, if you click Validate from within a ruleset, the list of rules shown includes rules that are not part of the selected ruleset. To work around this, manually select the rules that you want to use for the validation.
- PK12 Key/Cert pairs created with legacy algorithms are not supported for use with the
apic-gw-service
- As a result of increased security requirements in IBM
API Connect 10.0.8.0, if you
configured your non-container gateways to use a PK12 cert/key pair that was generated with legacy
algorithms, the Gateway service will fail to start.
Logs will show the following message:
[0x88e00011][apic-gw-service][error] apic-gw-service(default): API Connect Gateway Service caught unhandled rejection: Error: All sentinels are unreachable. Retrying from scratch after 10ms. Last error: unsupported
To resolve this problem, you must create a new key/cert pair and update your
apic-gw-service
object with this configuration. This can be done directly through the DataPower crypto tools, or through a third-party tool such as OpenSSLv3. note that OpenSSLv1 cannot be used because defaults to the legacy algorithms. - Gateway pod might show as not ready after upgrade
- Gateway pods might fail to display as READY after upgrading to IBM
API Connect 10.0.8.0. The pods
will show as RUNNING, but 0/1 READY, as in the following example:
$ kubectl get pods -n <namespace> NAME READY STATUS RESTARTS AGE ... gwv6-0 1/1 Running 0 1h gwv6-1 1/1 Running 0 1h gwv6-2 0/1 Running 0 1h ...
To work around this behavior, use the following command delete the pod that is not ready:
$ kubectl delete pod gwv6-2 -n <namespace>
If there are multiple gateway pods in the cluster, the delete operation will not result in any service disruption.
- Gateway services that span data centers
-
If you want to configure a gateway service that consists of gateways in different data centers, then note the following points:
- This configuration should be used only if TMS is required and an external OAuth provider cannot be used.
- Network latency between sites must be less than 30 ms.
- For the analytics service to span data centers, network latency between data centers must be less than 10 ms. If you want to associate an analytics service with your gateway service, and cannot meet this latency requirement, then choose one of the data centers to host the analytics service.
- Cloud Pak for Integration: Cannot add an API to the Asset Repository
- If you are using API Connect as a component of Cloud Pak for Integration, you cannot add an API to the Asset Repository from API Connect.
- Cloud Pak for Integration: Problem importing from Asset repository when user remains inactive
- If there is no user activity on the page while importing an API from the Cloud Pak for
Integration Asset repository, the user might see the following error:
could not add asset, parent window missing
Work around the issue by reloading the Import API page and then opening the Asset repository again.
- Limitations to the updated schema editor for API definitions
- The Definitions section of the API editor for the API
Manager and API Designer UIs was updated in
API Connect 10.0.6.0. However, the
OneOf
,AllOf
, andEnum
schema structures aren't handled properly by the UI. You can work around this issue by editing the source YAML of your API document. - API Designer with Local Test Environment (LTE) fails to log in using https://localhost with the error message "Incorrect username, password, or credentials"
- If you use the API Designer with the Local Test Environment and attempt to log in using the
localhost, the login fails. You can work around the issue by configuring API Designer credentials to
with the local host. Complete the following steps:
- Download and extract API Designer, then install the credentials file as explained in steps 1, 2, and 6 in Installing the toolkit.
- Edit the designer_credentials.json file and configure the following settings:
"endpoint": "https://localhost"
"manager_endpoint": "https://localhost/manager"
"client_id": Client Id
The client ID displays on the console when you start the LTE platform-apic-lte. For more information, see Testing an API with the Local Test Environment.
"client_secret": Client Secret
The client secret displays on the console when you start the LTE platform-apic-lte. For more information, see Testing an API with the Local Test Environment.
- Start API Designer and log in with the LTE using https://localhost as the Host URL (the management endpoint).
- Gateway timeout when validating very large API documents with the API governance service
- If you are validating very large API documents (5 MB, or 100,000 lines, or more), with all of
the rulesets selected by default, the API governance service is likely to take longer than 60
seconds to run, and this can result in an HTTP 504 gateway timeout. To workaround this issue, you
can take the following actions:
- Validate the API document by using only one ruleset at a time.
- Increase your component deployment profile, for example to a three replica profile. For more information, see Deployment and component profiles, and Planning your deployment topology and profiles on VMware.
- New ibm-apiconnect-operator pod fails to start due existing terminating ibm-apiconnect-operator pod holding lock
- The IBM APIConnect operator uses leader election (leader-for-life) to ensure that only one
operator/controller pod is running/active at a time. This relies on Kubernetes garbage collection to
clean up the lock when the old operator pod is deleted. The new ibm-apiconnect operator pod fails to
start if an existing terminating ibm-apiconnect operator pod is still holding the lock.
If you suspect you are encountering this issue, verify it by running the following command:
kubectl -n <namespace> logs <new-ibm-apiconnect-pod-name> | grep "LockOwner"
If the lock is the problem, the log contains a message similar to the following example:
existing lock","LockOwner":"ibm-apiconnect-xxxxxx held by terminating pod
Resolve the issue by completing one of the following options:
- Force delete the terminating pod by running the following command:
kubectl -n <namespace> delete pod <terminating-ibm-apiconnect-pod-name> --force --grace-period=0
Kubernetes deletes the lock when the pod is deleted.
- Remove the lock manually by running the following
command:
kubectl delete cm ibm-apiconnect-lock
- Force delete the terminating pod by running the following command:
- API Designer might hang while activating a large imported API
- When you import a large API using API Designer and attempt to activate the API in the import
wizard, the process might hang. If you encounter this issue, you can work around the problem by
completing the following steps:
- On your local file system, locate the autoproduct file called API-NAME-auto-product_API-VERSION.yaml.
- Delete the file.
- In API Designer, edit the newly imported API and activate it by clicking the Online toggle.
In general, it is best practice to activate an API using the Online toggle or by publishing the API with the Publish option.
- API Designer on Windows: APIs that use WSDL might encounter errors, or fail to activate, publish, or update.
- If you activate, publish, or update a REST or SOAP API that uses a WSDL file, the operation might fail and never complete. Work around the issue by using the autopublish API feature in the API editor.
- Permission restrictions in the API Designer UI
- The API Designer UI
currently has the following permission restrictions:
- Developers that have been given only View permissions cannot see the Test tab in the API editor. For developers to be able to see the Test tab, they must be given a different permission level. For information about the default permission levels available, see API Connect user roles.
- Users with API-Drafts permissions, but who don't have any Sandbox Catalog permissions, cannot see the Test preferences in the Sandbox Catalog. For these users to be able to see the Test preferences, they must be given the Developer or Administrator role on the Sandbox Catalog.
- Deleted security requirements might remain in the API source
- Security requirements that are deleted from APIs in the API Designer and API
Manager UIs, might still
remain in the source. To work around this issue, click the Source icon
in the API editor, and manually remove the security requirements.
- Sometimes the completion of an SFTP restore on the Management subsystem does not restore the actual data.
- There is a known issue where sometimes a restore from the Management subsystem's SFTP backup succeeds but the data is not restored. If this happens, run the restore again.
- Customized application view is reset to the default after upgrade if forums are also disabled
- When upgrading API Connect version 10.0.5.0
or earlier to version 10.0.5.1 or later, if you have disabled forums on your Developer Portal
site, and you have customized the application view, the customized application view will be reset to
the default after the upgrade. To work around this issue, you can save the customized application
view before running the upgrade, and then import the view into the upgraded Developer Portal.To save the customized application view, run the following commands:
Wheremkdir /tmp/tmp_config_<site-alias> drush @<site-alias> config-get views.view.applications > /tmp/tmp_config_<site-alias>/views.view.applications.yaml
<site-alias>
is the alias of your Developer Portal site.To import the saved view after the upgrade, put the views.view.applications.yaml file inside a directory where it's the only file, and then run the following command:drush @<site-alias> config-import --partial --source=/tmp/tmp_config_<site-alias>/
- Cannot add comments to APIs by using the Source view in the API Designer and API Manager UIs
- Comments cannot be added to APIs in the API Designer and API
Manager UIs by clicking the
Source icon
, and using the hashtag symbol.
- Changing Product visibility from Custom to Public doesn't automatically remove the Consumer organizations and groups
- In the API Designer and API Manager UIs, changing Product visibility from Custom to Public doesn't automatically remove the consumer organizations and groups, so the Product publish will fail. To work around this issue, manually remove all of the consumer organizations and groups.
- Stale cache can result in unexpected behavior in the Cloud Manager and API Manager UI
- Having a stale cache in your browser can result in unexpected behavior in the Cloud Manager
and API
Manager UI, such
as fetch errors, incorrect data being displayed, and blank pages. To work around this issue, particularly if you have just upgraded your deployment, complete the
following actions:
- Reload the browser window.
- If there is still an issue, clear the browser cache and then log back in to the UI.
- Try using a private browser window.
- If possible, try a different browser type.
- The "Select compatible gateway services" option in Test Preferences causes a "404 POST undefined" error while testing the API
- In the Select compatible gateway
services and choosing a specific gateway results in a "404 POST undefined" error when
you test the API in the API Explorer or the Test tab.
Workaround: To avoid this problem, select the Use all compatible gateway services target gateway option instead.
setting, selecting - Analytics command restrictions
- The following
--mode analytics
commands work only when the flag--return_format
is set to eitherjson
oryaml
:clustermgmt:catAllocation
clustermgmt:catIndices
clustermgmt:catNodes
clustermgmt:catRecovery
clustermgmt:catShards
clustermgmt:catAliases
text/plain
:clustermgmt:getNodesHotThreads
clustermgmt:getNodesHotThreadsById
- Options menus in the Catalog might be hidden
- In a Catalog, in any of the different tabs such as Consumers, or
Subscriptions, when clicking on the options icon
the menu items might be hidden. To work around the issue, reload the page and the menu items will appear.
- Override plan rate limits are not displayed in the Endpoint tab
- Any override plan rate limits that have been added to your API for individual operations are not displayed in the API Endpoint tab in the UI. Only the plan rate limit is displayed.
- Consumer rate limit notifications are not available in Version 10
- The ability to configure notifications for applications so that API consumers are alerted when the usage of an API is nearing its rate limit is not available.
- Configuration update mistakes cause admission webhook errors in OpenShift and Cloud Pak for Integration
- In OpenShift and Cloud Pak for Integration, configuration update errors can cause the
APIConnectCluster
instance to go into the Pending state with an underlying component throwing an admission webhook error.oc get apiconnectcluster -n APIC_namespace NAME READY STATUS VERSION RECONCILED VERSION AGE m1 admission webhook "vmanagementcluster.kb.io" denied the request: ManagementCluster.management.apiconnect.example.com "m1-mgmt" is invalid: [spec.databaseBackup.restartDB.accept: Invalid value: false: databaseBackup protocol has changed (from '' to 'objstore') and requires database restart. accept restartDB to restart database, spec.databaseBackup.restartDB.accept: Invalid value: false: databaseBackup configuration has changed and requires database restart. accept restartDB to restart database] Pending 10.0.1.5-3388-eus 10.0.1.5-3388-eus 2d6h
- The 'strict' SameSite cookie causes incorrect invitations to consumer organizations
- Using the 'strict' SameSite cookie might cause invitation links from emails to
send users to a registration page where they are asked to create a new consumer organization instead
of joining the organization they were invited to.
The workaround is to using the 'Lax' SameSite Cookie attribute.
- Limitations to the OpenAPI 3.0 support
- IBM API Connect supports the OpenAPI 3.0 specification, with some limitations. For information about what is supported, see OpenAPI 3.0 support.
- A GraphQL API that contains a
graphql-input-type-cost
rate limit fails to publish - A GraphQL API created in releases
earlier than IBM
API Connect
Version 10.0.3.0 might contain a
graphql-input-type-cost
rate limit, which is no longer supported. If you attempt to use the automatic activation mechanism to publish the API, or manually add the API to a Product and attempt to publish that Product, the publish operation will fail. You can resolve this problem in either of the following ways:- Remove the rate limit definition from the OpenAPI source for the API. For example, if the source
is in YAML format, remove the following lines:
- name: graphql-input-type-cost operation: consume
- Edit the source for the Product and define a
graphql-input-type-cost
rate limit in all of the Plans that include the API.Note: You can edit only a manually created Product, not a Product that is generated by the automatic activation mechanism.
- Remove the rate limit definition from the OpenAPI source for the API. For example, if the source
is in YAML format, remove the following lines:
- Update WSDL option is not supported in the API Designer UI
- The Update WSDL option to update the configuration of an existing SOAP API from a WSDL file or a .zip archive, is supported only in the API Manager UI, or by using the developer toolkit. This option is not supported in the API Designer UI.
- A Product republish, with the Preserve Subscriptions option, fails after a consumer organization group has been removed from the visibility settings
- If you remove a consumer organization group from the custom visibility settings for a Product, and that group contains a consumer organization with an application that is subscribed to the Product, an attempt to republish the Product with the Preserve Subscriptions option will fail even if that consumer organization is then added to the custom visibility settings individually.
- For a secured GraphQL API, GraphQL subscriptions cannot be tested by using the Test tab in the user interfaces
- For a GraphQL API that is secured by client ID, GraphQL subscriptions
cannot be tested by using the Test tab in the API Designer or API
Manager user interfaces. The
API can still be published and used in production.You can test GraphQL subscriptions in either of the following alternative ways:
- Remove client ID security from the API, for testing purposes, then use the Test tab.
- Use an external test tool.
- A published API whose assembly contains a catch block with an invalid policy will now correctly fail to republish
- Previously, policies in an assembly catch block were not validated, so if an incorrect policy configuration was coded in the OpenAPI source for an API, within an assembly catch block, the API would still publish successfully. Now, policies in an assembly catch block are validated, so such an API will fail to republish and will first need to be corrected.
- An existing user cannot be added to a Space
- If you attempt to add an existing user
to a Space, the operation cannot be completed because the Create button is
not enabled. Instead, use the invitation mechanism. For details, see Managing Space membership.Note: The invited user must use the Sign in option on the activation form, rather than completing the registration details and using Sign up.
- Simultaneous editing of the same API by multiple users might result in changes being overwritten
- If one user saves changes to an OpenAPI 3.0 API, another user who has the same API open in the API editor might have their changes overwritten.
- An API developed for IBM
API Connect Version 2018.4.1 that
uses an
apim.getvariable('message.body')
function call fails in IBM API Connect Version 10 - With the DataPower® API Gateway in IBM
API Connect Version 2018.4.1, the
type of object returned, for XML payloads, by an
apim.getvariable('context_name.body')
function call in a GatewayScript policy in an API assembly depends on how the variable is stored in the gateway context, as follows:- If the variable is stored as a Buffer (because the variable data was written as an XML string),
an XML Nodelist is returned provided that
contextname.headers.content-type
is an XML type. - If the variable is stored as a parsed XML Document (because the variable data was written as parsed XML, either as an XML document or an XML Nodelist), an XML document is returned.
With the DataPower API Gateway in IBM API Connect Version 10 however, the same function call always returns an XML Nodelist provided that
contextname.headers.content-type
is an XML type. Therefore, such an API developed for Version 2018.4.1 will fail in Version 10 if it is configured to expect an XML document, and will need to be reconfigured appropriately.The problem does not occur with such an API migrated from IBM API Connect Version 5 because, with that release,
apim.getvariable('context_name.body')
also returns an XML Nodelist for XML payloads, nor does it occur if you are using the DataPower Gateway (v5 compatible).context_name
could bemessage
,request
, or the name of the output from aninvoke
policy. - If the variable is stored as a Buffer (because the variable data was written as an XML string),
an XML Nodelist is returned provided that
- Validate policy limitations when GraphQL response contains a GraphQL server error
- When a GraphQL response contains a GraphQL server error and no
data, the assembly Validate policy generates an error on the missing data and overwrites the
payload. When the response contains partial data and an error, the assembly Validate action
validates the data and overwrites the payload. To work around this limitation, use the condition
$not($exists(message.body.errors))
in an assembly switch condition to skip the assembly Validate policy when the response contains errors.
- An OpenAPI definition that contains regular expression syntax fails validation
- IBM
API Connect supports the GO regular expression
syntax. When you import an OpenAPI definition into the API Designer or API
Manager user interfaces, or
validate one with the
apic validate
, the validation will fail if the OpenAPI source contains unsupported regular expression syntax, with errors that includeDoes not match format 'regex'
; for example:- Must validate at least one schema (anyOf) (context: (root).paths./example/types.post.parameters.0.schema.properties.items, line: 0, col: 0) - Must validate one and only one schema (oneOf) (context: (root).paths./example/types.post.parameters.0, line: 46164, col: 21) - paths./example/types.post.parameters.0.schema.properties.items.properties.pattern Does not match format 'regex' (context: (root).paths./example/types.post.parameters.0.schema.properties.items.properties.pattern, line: 0, col: 0)
- When scaling down a Kubernetes cluster of DataPower gateways, the cluster might enter an unrecoverable state if the peering primary instance is lost
- If you are scaling down a Kubernetes cluster of DataPower
gateways, you must first complete the following steps to ensure that the peering primary instance is
not lost:
- Incorrect UI behavior if API changes are not saved before creating a new API version
- If you make changes to an API definition and then attempt to create a new API version without first saving your changes, you are not prompted to save your changes until after you have completed the new version creation operation. If you click OK in the prompt, your new version is created but your original changes are lost; to create your new version and retain your original changes, you should click Cancel in the prompt, then click Save to save your original changes.
- Truststore and keystore settings in a shared TLS client profile are not reflected in API Manager
- If you create a shared TLS client profile in the Cloud Manager user interface and assign a truststore and a keystore, these truststore and keystore settings are not reflected in that TLS client profile in the API Manager user interface and cannot be set there. Furthermore, the TLS client profile list views display the value shared for the truststore and keystore rather than the correct values. To workaround this problem, create the truststore, keystore, and TLS client profile separately for each provider organization in the API Manager user interface, then enable the TLS client profile in your Catalogs as required.
- Login to the Cloud Manager or API Manager user interfaces might fail with error 431 if the browser has a large number of cookies
- Attempts to login to the Cloud Manager or API Manager user interfaces might fail if the HTTP header or cookie size is larger than 32 KB, a limitation that is imposed for security reasons. To resolve this problem, either clear the browser cache and cookies, or open a private window, then retry.
- The
apic cloud-settings:topology
command doesn't return a topology object - If you use the
apic cloud-settings:topology
command without a--format
parameter, it returns only the stringCloudTopology
rather than the cloud topology. To retrieve the topology object, include the--format
parameter, with a value of eitheryaml
orjson
as required. - The
apic cloud-settings:mail-server-configured
command response is incorrect - If you use the
apic cloud-settings:mail-server-configured
without a--format
parameter, it returns only the stringMailServerConfigured
rather than a boolean value to indicate whether an email server has been configured in the cloud settings. To retrieve the correct response, include the--format
parameter, with a value of eitheryaml
orjson
as required. - API Designer user interface might not open correctly if there are badly formed OpenAPI YAML files in the startup directory
- If, when launching the API Designer user interface, the
startup directory that you select contains one or more badly formed OpenAPI YAML files, the
interface might fail to load correctly, with an error such as the
following:
Cannot read property 'type' of null
To resolve this problem, remove any invalid OpenAPI YAML files from the startup directory, then restart the API Designer user interface.
- Pagination setting is global across the API Connect user interfaces
- If you set the Items per page value on any page in the Cloud Manager or API Manager user interface, that setting is then applied to all pages in both user interfaces in the same browser session. If you want to set the value separately for a specific page, open it in a private browser window. Such a setting in a private browser window is specific to that window and is lost when the window is closed.
- Setting resource limits in Kubernetes can cause containers to behave oddly at run time
- In Kubernetes, you can set resource limits for each container. However, limiting CPU means that the container processes might slow down if they try to use too much. Limiting memory means that the Kubernetes code kills processes that try to allocate memory that would take the container over its limit. These limits can result in odd behavior such as the containers continually trying to start a daemon process, such as nginx or node, and exiting with success immediately.
- Field validation for the Client security policy is incorrect
- When configuring a Client Security policy in an API
assembly in the API Designer
or API
Manager user
interfaces, there is the following incorrect validation behavior:
- The ID Name field is required, but the API definition can be saved without entering a value in the field.
- The Secret Name field is required only when the Secret Required option is selected, but the user interface indicates that the Secret Name field is required regardless. Furthermore, when the field is required, the API definition can be save without entering a value.
- If the Authenticate Client Method is set to Third party, the User Registry Name field is required, but the API definition can be saved without entering a value in this field.
- Requests that take longer than the value of the nginx timeout cause a 409 error
-
Requests that take longer than the value of the
nginx
timeout can cause a 409 error, such as:
The error occurs because the ingress controller is sending a repeat request when the previous one timed out. Your request will still continue processing in the background.Another request is operating on portal-subsystem-03729230-4df7-4419-94c9-e7691b616382 with the same name, please try again.
- Adding members to a provider organization might fail
- When adding members to a provider organization, the Cloud Manager user interface lists the current owner and members as well. If existing members are selected, the add action will fail.
- Upgrading with new certificates breaks analytics ingestion
- If the analytics subsystem certificates are modified after installation (by using the apicup certs set command , followed by the apicup install command to deploy the change to the cluster), the analytics ingestion ingress will no longer be recognized by the Gateway. The analytics service must be re-registered in the Cloud Manager interface so that the new certificates are recognized.
- User interfaces are not supported in Microsoft Edge
- The API Manager and Cloud Manager user interfaces are not supported in the Microsoft Edge web browser. To work in the user interfaces, use a different browser.
- New gateways cannot connect to existing gateways after the gateway peering certificates are changed
- Changing the gateway peering SSL certificates after gateway installation causes new gateways to be unable to connect to existing gateways, resulting in an outage. A full restart of the gateway StatefulSet (by deleting all the StatefulSet pods) is required for the gateways to come back online.
- No option to bulk delete APIs or Products
- In the API Designer and API Manager user interfaces, there is currently no option to delete multiple APIs or Products in a single operation; in the user interfaces, APIs and Products must be deleted individually. However, you can bulk delete APIs and Products by using the REST API or CLI interfaces.
- Cannot publish an API that has duplicate security definition entries
- The API Designer and API Manager user interfaces allow you to add duplicate security definitions to an API. However, attempts to publish the API will fail with OpenAPI validation errors. Ensure that the security definitions in an API are unique.
- Kubernetes might excessively evict pods
- In Kubernetes, if insufficient memory, CPU, or
storage resources are provided then pods are evicted randomly. To avoid this problem, allocate
sufficient resources to the pods. In addition, there are known issues with log rotation in
Kubernetes that are documented here: https://github.com/kubernetes/kubernetes/issues/59902. The
lack of log rotation can cause logs to grow, limited only by the amount of storage on the node,
eventually causing pods to restart. Therefore ensure that appropriate log levels are configured for
all API Connect
components.
You may also follow the instructions described here https://success.docker.com/article/how-to-setup-log-rotation-post-installation to set the maximum size of the log before it is rolled (max-size) and the maximum number of log files that can be present (max-files). More information about the Docker JSON File logging driver is available here: https://docs.docker.com/config/containers/logging/json-file/.
- Logging in to the API Connect user interfaces fails when using the Safari web browser
- If you are using the Safari web browser and a Basic Authorization header exists for the same DNS domain in which API Connect is running, attempts to log in to the API Connect user interfaces, or to sign up by using an activation link, fail. To avoid this problem, use an alternative web browser.
- Endpoints cannot be changed by using APICUP after they have been configured in Cloud Manager
-
The changing of endpoints with APICUP after they have been configured in Cloud Manager is not supported. Any such endpoint changes will not be propagated to a running deployment.
- An OAuth provider fails if the resources that it references aren't enabled in the Catalog
- If you enable an OAuth provider in a Catalog then any resources that it references, such as API user registries or TLS client profiles, must be enabled in the same Catalog; if not, then although the OAuth provider might publish successfully it will fail at run time. For information on enabling resources in a Catalog, see Creating and configuring Catalogs.
- Redact conditions within
switch
,operation-switch
, orif
policies might not execute after converting to DataPower API Gateway -
If an API Connect v5-compatible
redact
policy is found within aswitch
,operation-switch
, orif
policy, the migration utility does not move theredact
policy to the beginning or end of the assembly. The difference in the response between API Connect v5 and DataPower API Gateway may prevent data from being redacted in the DataPower API Gateway.For example, if an assembly includes a
switch
policy containing four redact conditions followed by aninvoke
policy, each redact condition redacts the response data. After porting to the API Gateway, the redact conditions remain inside theswitch
policy and target themessage.body
property for redaction. These redactions fail to execute because themessage.body
property has not yet been retrieved by theinvoke
policy. To correct this problem, you must move theinvoke
policy before theswitch
policy in the assembly. - Redact on invalid XPath fails after converting to DataPower API Gateway
Redact
on invalid XPath fails after converting to DataPower API Gateway. Application Management Unit (AMU) version 10.0.8.0-R0 support the conversion of the Redact policy, but only to Redact version 1.5.0, not 2.0.0.- Numerical Handling in YAML Configurations and Precision Constraints in API Manager UI
-
- In the API Manager UI YAML configurations, numbers in exponential notation (for example 1e20) are handled differently based on their exponent value. Numbers with an exponent less than or equal to 20 are converted to their full integer form (for example 1e20 becomes 100,000,000,000,000,000,000) for display and processing. Numbers with an exponent greater than 20 stay in exponential notation (for example 1e21). Numbers that exceed the range of integers that are supported by DataPower JSON schema validation (-9,007,199,254,740,992 to 9,007,199,254,740,992) result in validation errors or unexpected behavior.
- JavaScript's inherent precision limitations truncate floating-point numbers to approximately 17
significant digits. For example:
- Input: 0.123456789012345678901234567890
- Processed Value: 0.12345678901234568