Package org.apache.hadoop.yarn.server.resourcemanager.recovery

Class ZKRMStateStore

java.lang.Object

org.apache.hadoop.service.AbstractService

org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore

org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore

All Implemented Interfaces:

java.io.Closeable, java.lang.AutoCloseable, org.apache.hadoop.service.Service

@Private
@Unstable
public class ZKRMStateStore
extends RMStateStore

RMStateStore implementation backed by ZooKeeper. The znode structure is as follows: ROOT_DIR_PATH |--- VERSION_INFO |--- EPOCH_NODE |--- RM_ZK_FENCING_LOCK |--- RM_APP_ROOT | |----- HIERARCHIES | | |----- 1 | | | |----- (#ApplicationId barring last character) | | | | |----- (#Last character of ApplicationId) | | | | | |----- (#ApplicationAttemptIds) | | | .... | | | | | |----- 2 | | | |----- (#ApplicationId barring last 2 characters) | | | | |----- (#Last 2 characters of ApplicationId) | | | | | |----- (#ApplicationAttemptIds) | | | .... | | | | | |----- 3 | | | |----- (#ApplicationId barring last 3 characters) | | | | |----- (#Last 3 characters of ApplicationId) | | | | | |----- (#ApplicationAttemptIds) | | | .... | | | | | |----- 4 | | | |----- (#ApplicationId barring last 4 characters) | | | | |----- (#Last 4 characters of ApplicationId) | | | | | |----- (#ApplicationAttemptIds) | | | .... | | | | |----- (#ApplicationId1) | | |----- (#ApplicationAttemptIds) | | | |----- (#ApplicationId2) | | |----- (#ApplicationAttemptIds) | .... | |--- RM_DT_SECRET_MANAGER_ROOT |----- RM_DT_SEQUENTIAL_NUMBER_ZNODE_NAME |----- RM_DELEGATION_TOKENS_ROOT_ZNODE_NAME | |----- 1 | | |----- (#TokenId barring last character) | | | |----- (#Last character of TokenId) | | .... | |----- 2 | | |----- (#TokenId barring last 2 characters) | | | |----- (#Last 2 characters of TokenId) | | .... | |----- 3 | | |----- (#TokenId barring last 3 characters) | | | |----- (#Last 3 characters of TokenId) | | .... | |----- 4 | | |----- (#TokenId barring last 4 characters) | | | |----- (#Last 4 characters of TokenId) | | .... | |----- Token_1 | |----- Token_2 | .... | |----- RM_DT_MASTER_KEYS_ROOT_ZNODE_NAME | |----- Key_1 | |----- Key_2 .... |--- AMRMTOKEN_SECRET_MANAGER_ROOT |----- currentMasterKey |----- nextMasterKey |-- RESERVATION_SYSTEM_ROOT |------PLAN_1 | |------ RESERVATION_1 | |------ RESERVATION_2 | .... |------PLAN_2 .... |-- PROXY_CA_ROOT |----- caCert |----- caPrivateKey Note: Changes from 1.1 to 1.2 - AMRMTokenSecretManager state has been saved separately. The currentMasterkey and nextMasterkey have been stored. Also, AMRMToken has been removed from ApplicationAttemptState. Changes from 1.2 to 1.3, Addition of ReservationSystem state. Changes from 1.3 to 1.4 - Change the structure of application znode by splitting it in 2 parts, depending on a configurable split index. This limits the number of application znodes returned in a single call while loading app state. Changes from 1.4 to 1.5 - Change the structure of delegation token znode by splitting it in 2 parts, depending on a configurable split index. This limits the number of delegation token znodes returned in a single call while loading tokens state.

Nested Class Summary

Nested classes/interfaces inherited from class org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore

RMStateStore.ProxyCAState, RMStateStore.RMDTSecretManagerState, RMStateStore.RMState, RMStateStore.RMStateStoreState

Nested classes/interfaces inherited from interface org.apache.hadoop.service.Service

org.apache.hadoop.service.Service.STATE

Field Summary

Fields

Modifier and Type	Field	Description
protected static org.apache.hadoop.yarn.server.records.Version	CURRENT_VERSION_INFO
protected int	delegationTokenNodeSplitIndex
protected ZKRMStateStoreOpDurations	opDurations
static java.lang.String	RM_APP_ROOT_HIERARCHIES
static java.lang.String	ROOT_ZNODE_NAME
protected java.lang.String	znodeWorkingPath

Fields inherited from class org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore

AM_CLIENT_TOKEN_MASTER_KEY_NAME, AM_RM_TOKEN_SERVICE, AMRMTOKEN_SECRET_MANAGER_ROOT, baseEpoch, DELEGATION_KEY_PREFIX, DELEGATION_TOKEN_PREFIX, DELEGATION_TOKEN_SEQUENCE_NUMBER_PREFIX, EPOCH_NODE, PROXY_CA_CERT_NODE, PROXY_CA_PRIVATE_KEY_NODE, PROXY_CA_ROOT, RESERVATION_SYSTEM_ROOT, resourceManager, RM_APP_ROOT, RM_DT_SECRET_MANAGER_ROOT, rmStateStoreEventHandler, VERSION_NODE

Constructor Summary

Constructors

Constructor	Description
ZKRMStateStore()

Method Summary

Modifier and Type	Method	Description
protected void	closeInternal()	Derived classes close themselves using this method.
protected java.util.List<org.apache.zookeeper.data.ACL>	constructZkRootNodeACL(org.apache.hadoop.conf.Configuration conf, java.util.List<org.apache.zookeeper.data.ACL> sourceACLs)	Given the Configuration and ACLs used (sourceACLs) for ZooKeeper access, construct the ACLs for the store's root node.
void	deleteStore()	Derived classes must implement this method to delete the state store.
long	getAndIncrementEpoch()	Get the current epoch of RM and increment the value.
protected org.apache.hadoop.yarn.server.records.Version	getCurrentVersion()	Get the current version of the underlying state store.
void	initInternal(org.apache.hadoop.conf.Configuration conf)	Derived classes initialize themselves using this method.
RMStateStore.RMState	loadState()	Blocking API The derived class must recover state from the store and return a new RMState object populated with that state This must not be called on the dispatcher thread.
protected org.apache.hadoop.yarn.server.records.Version	loadVersion()	Derived class use this method to load the version information from state store.
void	removeApplication(org.apache.hadoop.yarn.api.records.ApplicationId removeAppId)	Derived classes must implement this method to remove application from the state store.
protected void	removeApplicationAttemptInternal(org.apache.hadoop.yarn.api.records.ApplicationAttemptId appAttemptId)	Blocking API Derived classes must implement this method to remove the state of specified attempt.
protected void	removeApplicationStateInternal(ApplicationStateData appState)	Blocking API Derived classes must implement this method to remove the state of an application and its attempts.
protected void	removeReservationState(java.lang.String planName, java.lang.String reservationIdName)	Blocking API Derived classes must implement this method to remove the state of a reservation allocation.
protected void	removeRMDelegationTokenState(org.apache.hadoop.yarn.security.client.RMDelegationTokenIdentifier rmDTIdentifier)	Blocking API Derived classes must implement this method to remove the state of RMDelegationToken.
protected void	removeRMDTMasterKeyState(org.apache.hadoop.security.token.delegation.DelegationKey delegationKey)	Blocking API Derived classes must implement this method to remove the state of DelegationToken Master Key.
void	startInternal()	Derived classes start themselves using this method.
protected void	storeApplicationAttemptStateInternal(org.apache.hadoop.yarn.api.records.ApplicationAttemptId appAttemptId, ApplicationAttemptStateData attemptStateDataPB)	Blocking API Derived classes must implement this method to store the state of an application attempt.
void	storeApplicationStateInternal(org.apache.hadoop.yarn.api.records.ApplicationId appId, ApplicationStateData appStateDataPB)	Blocking API Derived classes must implement this method to store the state of an application.
protected void	storeOrUpdateAMRMTokenSecretManagerState(AMRMTokenSecretManagerState amrmTokenSecretManagerState, boolean isUpdate)	Blocking API Derived classes must implement this method to store or update the state of AMRMToken Master Key.
protected void	storeProxyCACertState(java.security.cert.X509Certificate caCert, java.security.PrivateKey caPrivateKey)	Blocking API Derived classes must implement this method to store the CA Certificate and Private Key.
protected void	storeReservationState(org.apache.hadoop.yarn.proto.YarnProtos.ReservationAllocationStateProto reservationAllocation, java.lang.String planName, java.lang.String reservationIdName)	Blocking API Derived classes must implement this method to store the state of a reservation allocation.
protected void	storeRMDelegationTokenState(org.apache.hadoop.yarn.security.client.RMDelegationTokenIdentifier rmDTIdentifier, java.lang.Long renewDate)	Blocking API Derived classes must implement this method to store the state of RMDelegationToken and sequence number.
protected void	storeRMDTMasterKeyState(org.apache.hadoop.security.token.delegation.DelegationKey delegationKey)	Blocking API Derived classes must implement this method to store the state of DelegationToken Master Key.
protected void	storeVersion()	Derived class use this method to store the version information.
protected void	updateApplicationAttemptStateInternal(org.apache.hadoop.yarn.api.records.ApplicationAttemptId appAttemptId, ApplicationAttemptStateData attemptStateDataPB)
protected void	updateApplicationStateInternal(org.apache.hadoop.yarn.api.records.ApplicationId appId, ApplicationStateData appStateDataPB)
protected void	updateRMDelegationTokenState(org.apache.hadoop.yarn.security.client.RMDelegationTokenIdentifier rmDTIdentifier, java.lang.Long renewDate)	Blocking API Derived classes must implement this method to update the state of RMDelegationToken and sequence number.

Methods inherited from class org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore

checkVersion, getCredentialsFromAppAttempt, getRMStateStoreEventHandler, getRMStateStoreState, handleStoreEvent, isFencedState, nextEpoch, notifyStoreOperationFailed, removeApplication, removeApplicationAttempt, removeReservation, removeRMDelegationToken, removeRMDTMasterKey, serviceInit, serviceStart, serviceStop, setResourceManager, setRMDispatcher, storeNewApplication, storeNewApplicationAttempt, storeNewReservation, storeOrUpdateAMRMTokenSecretManager, storeProxyCACert, storeRMDelegationToken, storeRMDTMasterKey, updateApplicationAttemptState, updateApplicationState, updateApplicationState, updateApplicationStateSynchronously, updateFencedState, updateRMDelegationToken

Methods inherited from class org.apache.hadoop.service.AbstractService

close, getBlockers, getConfig, getFailureCause, getFailureState, getLifecycleHistory, getName, getServiceState, getStartTime, init, isInState, noteFailure, putBlocker, registerGlobalListener, registerServiceListener, removeBlocker, setConfig, start, stop, toString, unregisterGlobalListener, unregisterServiceListener, waitForServiceToStop

Methods inherited from class java.lang.Object

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

Field Detail

ROOT_ZNODE_NAME

@VisibleForTesting
public static final java.lang.String ROOT_ZNODE_NAME

See Also:

Constant Field Values

CURRENT_VERSION_INFO

protected static final org.apache.hadoop.yarn.server.records.Version CURRENT_VERSION_INFO

RM_APP_ROOT_HIERARCHIES

@VisibleForTesting
public static final java.lang.String RM_APP_ROOT_HIERARCHIES

See Also:

Constant Field Values

znodeWorkingPath

@VisibleForTesting
protected java.lang.String znodeWorkingPath

delegationTokenNodeSplitIndex

@VisibleForTesting
protected int delegationTokenNodeSplitIndex

opDurations

@VisibleForTesting
protected ZKRMStateStoreOpDurations opDurations

Constructor Detail

ZKRMStateStore

public ZKRMStateStore()

Method Detail

constructZkRootNodeACL

@VisibleForTesting
@Private
@Unstable
protected java.util.List<org.apache.zookeeper.data.ACL> constructZkRootNodeACL(org.apache.hadoop.conf.Configuration conf,
                                                                               java.util.List<org.apache.zookeeper.data.ACL> sourceACLs)
                                                                        throws java.security.NoSuchAlgorithmException

Given the Configuration and ACLs used (sourceACLs) for ZooKeeper access, construct the ACLs for the store's root node. In the constructed ACL, all the users allowed by sourceACLs are given read-write-admin access, while the current RM has exclusive create-delete access. To be called only when HA is enabled and the configuration doesn't set an ACL for the root node.

Parameters:

conf - the configuration

sourceACLs - the source ACLs

Returns:

ACLs for the store's root node

Throws:

java.security.NoSuchAlgorithmException - thrown if the digest algorithm used by Zookeeper cannot be found

initInternal

public void initInternal(org.apache.hadoop.conf.Configuration conf)
                  throws java.io.IOException,
                         java.security.NoSuchAlgorithmException

Description copied from class: RMStateStore

Derived classes initialize themselves using this method.

Specified by:

initInternal in class RMStateStore

Parameters:

conf - Configuration.

Throws:

java.io.IOException

java.security.NoSuchAlgorithmException

startInternal

public void startInternal()
                   throws java.lang.Exception

Description copied from class: RMStateStore

Derived classes start themselves using this method. The base class is started and the event dispatcher is ready to use at this point.

Specified by:

startInternal in class RMStateStore

Throws:

java.lang.Exception - error occur.

closeInternal

protected void closeInternal()
                      throws java.lang.Exception

Description copied from class: RMStateStore

Derived classes close themselves using this method. The base class will be closed and the event dispatcher will be shutdown after this.

Specified by:

closeInternal in class RMStateStore

Throws:

java.lang.Exception - error occur.

getCurrentVersion

protected org.apache.hadoop.yarn.server.records.Version getCurrentVersion()

Description copied from class: RMStateStore

Get the current version of the underlying state store.

Specified by:

getCurrentVersion in class RMStateStore

Returns:

current version.

storeVersion

protected void storeVersion()
                     throws java.lang.Exception

Description copied from class: RMStateStore

Derived class use this method to store the version information.

Specified by:

storeVersion in class RMStateStore

Throws:

java.lang.Exception - error occur.

loadVersion

protected org.apache.hadoop.yarn.server.records.Version loadVersion()
                                                             throws java.lang.Exception

Description copied from class: RMStateStore

Derived class use this method to load the version information from state store.

Specified by:

loadVersion in class RMStateStore

Returns:

current version.

Throws:

java.lang.Exception - error occur.

getAndIncrementEpoch

public long getAndIncrementEpoch()
                          throws java.lang.Exception

Description copied from class: RMStateStore

Get the current epoch of RM and increment the value.

Specified by:

getAndIncrementEpoch in class RMStateStore

Returns:

current epoch.

Throws:

java.lang.Exception - error occur.

loadState

public RMStateStore.RMState loadState()
                               throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API The derived class must recover state from the store and return a new RMState object populated with that state This must not be called on the dispatcher thread.

Specified by:

loadState in class RMStateStore

Returns:

RMState.

Throws:

java.lang.Exception - error occur.

storeApplicationStateInternal

public void storeApplicationStateInternal(org.apache.hadoop.yarn.api.records.ApplicationId appId,
                                          ApplicationStateData appStateDataPB)
                                   throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to store the state of an application.

Specified by:

storeApplicationStateInternal in class RMStateStore

Parameters:

appId - application Id.

appStateDataPB - application StateData.

Throws:

java.lang.Exception - error occur.

updateApplicationStateInternal

protected void updateApplicationStateInternal(org.apache.hadoop.yarn.api.records.ApplicationId appId,
                                              ApplicationStateData appStateDataPB)
                                       throws java.lang.Exception

Specified by:

updateApplicationStateInternal in class RMStateStore

Throws:

java.lang.Exception

storeApplicationAttemptStateInternal

protected void storeApplicationAttemptStateInternal(org.apache.hadoop.yarn.api.records.ApplicationAttemptId appAttemptId,
                                                    ApplicationAttemptStateData attemptStateDataPB)
                                             throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to store the state of an application attempt.

Specified by:

storeApplicationAttemptStateInternal in class RMStateStore

Parameters:

appAttemptId - Application AttemptId.

attemptStateDataPB - Application AttemptStateData.

Throws:

java.lang.Exception - error occur.

updateApplicationAttemptStateInternal

protected void updateApplicationAttemptStateInternal(org.apache.hadoop.yarn.api.records.ApplicationAttemptId appAttemptId,
                                                     ApplicationAttemptStateData attemptStateDataPB)
                                              throws java.lang.Exception

Specified by:

updateApplicationAttemptStateInternal in class RMStateStore

Throws:

java.lang.Exception

removeApplicationAttemptInternal

protected void removeApplicationAttemptInternal(org.apache.hadoop.yarn.api.records.ApplicationAttemptId appAttemptId)
                                         throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to remove the state of specified attempt.

Specified by:

removeApplicationAttemptInternal in class RMStateStore

Parameters:

appAttemptId - application attempt id.

Throws:

java.lang.Exception - exception occurs.

removeApplicationStateInternal

protected void removeApplicationStateInternal(ApplicationStateData appState)
                                       throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to remove the state of an application and its attempts.

Specified by:

removeApplicationStateInternal in class RMStateStore

Parameters:

appState - ApplicationStateData.

Throws:

java.lang.Exception - error occurs.

storeRMDelegationTokenState

protected void storeRMDelegationTokenState(org.apache.hadoop.yarn.security.client.RMDelegationTokenIdentifier rmDTIdentifier,
                                           java.lang.Long renewDate)
                                    throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to store the state of RMDelegationToken and sequence number.

Specified by:

storeRMDelegationTokenState in class RMStateStore

Parameters:

rmDTIdentifier - RMDelegationTokenIdentifier.

renewDate - token renew date.

Throws:

java.lang.Exception - error occur.

removeRMDelegationTokenState

protected void removeRMDelegationTokenState(org.apache.hadoop.yarn.security.client.RMDelegationTokenIdentifier rmDTIdentifier)
                                     throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to remove the state of RMDelegationToken.

Specified by:

removeRMDelegationTokenState in class RMStateStore

Parameters:

rmDTIdentifier - RMDelegationTokenIdentifier.

Throws:

java.lang.Exception - error occurs.

updateRMDelegationTokenState

protected void updateRMDelegationTokenState(org.apache.hadoop.yarn.security.client.RMDelegationTokenIdentifier rmDTIdentifier,
                                            java.lang.Long renewDate)
                                     throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to update the state of RMDelegationToken and sequence number.

Specified by:

updateRMDelegationTokenState in class RMStateStore

Parameters:

rmDTIdentifier - RMDelegationTokenIdentifier.

renewDate - token renew date.

Throws:

java.lang.Exception - error occurs.

storeRMDTMasterKeyState

protected void storeRMDTMasterKeyState(org.apache.hadoop.security.token.delegation.DelegationKey delegationKey)
                                throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to store the state of DelegationToken Master Key.

Specified by:

storeRMDTMasterKeyState in class RMStateStore

Parameters:

delegationKey - DelegationToken Master Key.

Throws:

java.lang.Exception - error occur.

removeRMDTMasterKeyState

protected void removeRMDTMasterKeyState(org.apache.hadoop.security.token.delegation.DelegationKey delegationKey)
                                 throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to remove the state of DelegationToken Master Key.

Specified by:

removeRMDTMasterKeyState in class RMStateStore

Parameters:

delegationKey - DelegationKey.

Throws:

java.lang.Exception - exception occurs.

deleteStore

public void deleteStore()
                 throws java.lang.Exception

Description copied from class: RMStateStore

Derived classes must implement this method to delete the state store.

Specified by:

deleteStore in class RMStateStore

Throws:

java.lang.Exception - exception occurs.

removeApplication

public void removeApplication(org.apache.hadoop.yarn.api.records.ApplicationId removeAppId)
                       throws java.lang.Exception

Description copied from class: RMStateStore

Derived classes must implement this method to remove application from the state store.

Specified by:

removeApplication in class RMStateStore

Parameters:

removeAppId - application Id.

Throws:

java.lang.Exception - exception occurs.

storeOrUpdateAMRMTokenSecretManagerState

protected void storeOrUpdateAMRMTokenSecretManagerState(AMRMTokenSecretManagerState amrmTokenSecretManagerState,
                                                        boolean isUpdate)
                                                 throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to store or update the state of AMRMToken Master Key.

Specified by:

storeOrUpdateAMRMTokenSecretManagerState in class RMStateStore

Parameters:

amrmTokenSecretManagerState - amrmTokenSecretManagerState.

isUpdate - true, update; otherwise not update.

Throws:

java.lang.Exception - exception occurs.

removeReservationState

protected void removeReservationState(java.lang.String planName,
                                      java.lang.String reservationIdName)
                               throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to remove the state of a reservation allocation.

Specified by:

removeReservationState in class RMStateStore

Parameters:

planName - plan Name.

reservationIdName - reservationId Name.

Throws:

java.lang.Exception - exception occurs.

storeReservationState

protected void storeReservationState(org.apache.hadoop.yarn.proto.YarnProtos.ReservationAllocationStateProto reservationAllocation,
                                     java.lang.String planName,
                                     java.lang.String reservationIdName)
                              throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to store the state of a reservation allocation.

Specified by:

storeReservationState in class RMStateStore

Parameters:

reservationAllocation - reservation Allocation.

planName - plan Name.

reservationIdName - reservationId Name.

Throws:

java.lang.Exception - error occurs.

storeProxyCACertState

protected void storeProxyCACertState(java.security.cert.X509Certificate caCert,
                                     java.security.PrivateKey caPrivateKey)
                              throws java.lang.Exception

Description copied from class: RMStateStore

Blocking API Derived classes must implement this method to store the CA Certificate and Private Key.

Specified by:

storeProxyCACertState in class RMStateStore

Parameters:

caCert - X509Certificate.

caPrivateKey - PrivateKey.

Throws:

java.lang.Exception - error occurs.