edu.iu.dsc.tws.rsched.schedulers.k8s

Class PodWatchUtils

java.lang.Object

edu.iu.dsc.tws.rsched.schedulers.k8s.PodWatchUtils

public final class PodWatchUtils
extends java.lang.Object

this class is used to provide methods related to watching pods in a job

Method Summary

Modifier and Type	Method and Description
static void	close()
static java.util.HashMap<java.lang.String,java.lang.String>	discoverRunningPodIPs(java.util.ArrayList<java.lang.String> podNames, java.lang.String namespace, java.lang.String labelSelector, int timeout) watch pods until getting the Running event for all the pods in the given list return pod names and IP addresses as a HashMap
static ApiClient	getApiClient()
static CoreV1Api	getCoreApi()
static java.lang.String	getIpByWatchingPodToRunning(java.lang.String namespace, java.lang.String podName, int timeout) watch the given pod until it is Running and get its IP we assume that the pod is constructed as a StatefulSet
static java.lang.String	getJobMasterIpByWatchingPodToRunning(java.lang.String namespace, java.lang.String jobID, int timeout) watch the job master pod until it is Running and get its IP we assume that the job master has the unique twister2-role label and value pair
static java.lang.String	getNodeIP(java.lang.String namespace, java.lang.String jobID, java.lang.String podIP) get the IP of the node where the pod with that name is running
static java.util.ArrayList<java.lang.String>	getWorkerIPsByWatchingPodsToRunning(java.lang.String namespace, java.lang.String jobID, int numberOfPods, int timeout) watch the worker pods until they are Running and get their IP addresses we assume that workers have the unique twister2-role label and value pair we get the ip addresses of all workers including the worker pod calling this method
static void	testGetPodList(java.lang.String namespace) a test method to see whether kubernetes java client can connect to kubernetes master and get the pod list
static void	testWatchPods(java.lang.String namespace, java.lang.String jobID, int timeout) test watch pods method in the worker pod
static boolean	watchPodsToRunning(java.lang.String namespace, java.lang.String jobID, java.util.HashMap<java.lang.String,java.lang.Boolean> pods, int timeout) watch pods until getting the Running event for all the pods in the given map mark each pod that reached to Running state in the given map
static boolean	watchPodsToStarting(java.lang.String namespace, java.lang.String jobID, java.util.HashMap<java.lang.String,java.lang.Boolean> pods, int timeout) watch all pods in the given list until they become Starting flag the pods with a true value in the given HashMap

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

getCoreApi

public static CoreV1Api getCoreApi()

getApiClient

public static ApiClient getApiClient()

close

public static void close()

discoverRunningPodIPs

public static java.util.HashMap<java.lang.String,java.lang.String> discoverRunningPodIPs(java.util.ArrayList<java.lang.String> podNames,
                                                                                         java.lang.String namespace,
                                                                                         java.lang.String labelSelector,
                                                                                         int timeout)

watch pods until getting the Running event for all the pods in the given list return pod names and IP addresses as a HashMap

watchPodsToRunning

public static boolean watchPodsToRunning(java.lang.String namespace,
                                         java.lang.String jobID,
                                         java.util.HashMap<java.lang.String,java.lang.Boolean> pods,
                                         int timeout)

watch pods until getting the Running event for all the pods in the given map mark each pod that reached to Running state in the given map

watchPodsToStarting

public static boolean watchPodsToStarting(java.lang.String namespace,
                                          java.lang.String jobID,
                                          java.util.HashMap<java.lang.String,java.lang.Boolean> pods,
                                          int timeout)

watch all pods in the given list until they become Starting flag the pods with a true value in the given HashMap

getNodeIP

public static java.lang.String getNodeIP(java.lang.String namespace,
                                         java.lang.String jobID,
                                         java.lang.String podIP)

get the IP of the node where the pod with that name is running

testGetPodList

public static void testGetPodList(java.lang.String namespace)

a test method to see whether kubernetes java client can connect to kubernetes master and get the pod list

testWatchPods

public static void testWatchPods(java.lang.String namespace,
                                 java.lang.String jobID,
                                 int timeout)

test watch pods method in the worker pod

getIpByWatchingPodToRunning

public static java.lang.String getIpByWatchingPodToRunning(java.lang.String namespace,
                                                           java.lang.String podName,
                                                           int timeout)

watch the given pod until it is Running and get its IP we assume that the pod is constructed as a StatefulSet

getJobMasterIpByWatchingPodToRunning

public static java.lang.String getJobMasterIpByWatchingPodToRunning(java.lang.String namespace,
                                                                    java.lang.String jobID,
                                                                    int timeout)

watch the job master pod until it is Running and get its IP we assume that the job master has the unique twister2-role label and value pair

getWorkerIPsByWatchingPodsToRunning

public static java.util.ArrayList<java.lang.String> getWorkerIPsByWatchingPodsToRunning(java.lang.String namespace,
                                                                                        java.lang.String jobID,
                                                                                        int numberOfPods,
                                                                                        int timeout)

watch the worker pods until they are Running and get their IP addresses we assume that workers have the unique twister2-role label and value pair we get the ip addresses of all workers including the worker pod calling this method

getting IP addresses by list method does not work, since uninitialized pod IPs are not returned by list method

return null, if it can not get the full list