Skip to content

Latest commit

 

History

History
106 lines (71 loc) · 6.04 KB

README.md

File metadata and controls

106 lines (71 loc) · 6.04 KB
Raw

Kubernetes API prolog client

The prolog module and predicates for communicating with the kubernetes API server, with support of standard configuration options via KUBECONFIG, ~/.kube/config, and support for accessing the API from the k8s pods.

USAGE

Use the one of the exported predicates for CRUD operations or to watchi the resource lists. See Kubernetes API concepts for details about basic concept used in this module.

Example of usage:

?- use(library(k8s_client)).
?- k8s_get_resource(core, v1, pods, PodName, _, [k8s_namespace(myns)]).
PodName = "dex-567cdd88cd-dbv9x" ;
PodName = "envoy-proxy-599b679cf7-94764" ;

Exported predicates from the module k8s_client

k8s_create_resource(+ApiGroup:atom, +Version:atom, +ResourceTypeName:atom, +Instance:dict, -InstanceOut:dict, +Options:list) is semidet

Creates a resource at the Kubernetes API and unifies the InstanceOut with the server response. Options are same as for the predicate k8s_get_resource/6

k8s_delete_resource(+ApiGroup:atom, +Version:atom, +ResourceTypeName:atom, +Instance:dict/atom, +Options:list) is semidet

Deletes a resource at the Kubernetes API. Options are same as for the predicate k8s_get_resource/6

k8s_get_resource(+ApiGroup:atom, +Version:atom, +ResourceTypeName:atom, -InstanceName:atomic, -Instance:dict, +Options:list) is nondet

k8s_get_resource(+ApiGroup:atom, +Version:atom, +ResourceTypeName:atom, +InstanceName:atomic, -Instance:dict, +Options:list) is nondet

Unifies InstanceName - and Instance with the object representing resource of the kubernetes API. If the InstanceName is not bound then all instances are retrieved. ApiGroup is either the valid name of the Kubernetes API Group or the core atom.

The actual cluster address, context, and namespace is provided either options or loaded from the configuration. The below options are supported, in addition to options passed down to the http_open/3 predicate:

  • k8s_config(Config:dict) - kubectl configuration - if not provided then the configuation is loaded by first succesfull of the following possibilities

    • resolving and mergin config files specified by environment variable KUBECONFIG if existing;
    • loading ~/.kube/config file if existing;
    • using pod service account if the process is executed from the pod.

  • k8s_context(Context:atom) - name of the context to use from the k8s_config option. If not specified, then current-context property of k8s_config is used as a default value

  • k8s_namespace(Namespace) - the namespace from which to load resource instance(s). Shall be set to all if all namespaces shall be listed (will fail if InstanceName is bound and resource is namespaced). If not specified then the namespace provided as part of the context or the default namespace will be used.

  • k8s_resource_types_mode(Mode) where Mode is one of cache (default), renew, remote, local. Default to local. Used when resolving if the resource is namespaced. See also k8s_resource_types/2

  • k8s_selectors(Selectors:list) - list of selectors to apply for the resource retrieval(plain, not encoded form)

  • k8s_query(Query:term) - a query parameter to add to the REST call in form as specified by the predicate uri_query_components/2. This option is concatenated if used multiple times

k8s_resources(-ResourceTypes:list(dict), +Options) is semidet

Unifies the ResourceTypes with list of resource types available on the cluster. Primary used for resource discovery. Options are same as for the k8s_get_resource/6 with the extra option:

  • k8s_resource_types_mode(Mode) where Mode is one of cache (default), renew, remote, local. In the standard mode, the resources are cached per cluster and retrieved from the cache if available. If renew or remote mode is requested then the resource list is retrieved directly from the API server. The 'remote' and local mode leaves the cache untouched. The caching significantly speed up consequent operation with kubernetes resources

k8s_update_resource(+ApiGroup:atom, +Version:atom, +ResourceTypeName:atom, +Instanc:dict, -InstanceOut:dict, +Options:list) is semidet

Updates a resource at the Kubernetes API and unifies the InstanceOut with the server response. Options are same as for the predicate k8s_get_resource/6

k8s_watch_resources(:Callback, +ApiGroup:atom, +Version:atom, +ResourceTypeName:atom, +Options:list) is det

This predicate watches the changes of the resource list specified by the arguments and call call(:Goal, ChangeType:atom, ResourceInstance:dict) for each modification in the list. The ChangeType is one of the added, modified, deleted atom. Initial list of instances is provided as sequence of added callbacks after the call of this predicate.

The call is blocking the caller thread.

Options are same as for the k8s_get_resource/6 with the extra option:

  • k8s_resource_version(ResourceVersion:atom) - if specified the initial list is retrieved for the changed since the specified resource version. This option is used primary for internal purposes, and can be reset back to 0.

k8s_watch_resources_async(:Callback, +ApiGroup:atom, +Version:atom, +ResourceTypeName:atom, -StopWatcher, +Options:list) is det

Forks the separated background thread in which the predicate k8s_watch_resources/5 is executed.

The background thread can be finished and/or joined by calling call(StopWatcher, Mode:atom), where Mode is either join, or stop. The later mode will break the watching thread and force it to exit before joining it. If the Mode is unbound, then behavior is equivalent to the stop mode, and the Mode is boud to the exit status of the watcher thread.

Other arguments are same as for the k8s_watch_resources/5. Be aware that the Callback is invoked from the different thread than the thread calling this predicate.

Development

To debug the module, load the debug.pl file into prolog top.

Testing

Only manual tests of API calls executed