Package org.apache.storm.cluster
Class StormClusterStateImpl
java.lang.Object
org.apache.storm.cluster.StormClusterStateImpl
- All Implemented Interfaces:
IStormClusterState
-
Constructor Summary
ConstructorsConstructorDescriptionStormClusterStateImpl(IStateStorage stateStorage, ILocalAssignmentsBackend assignmentsassignmentsBackend, ClusterStateContext context, boolean shouldCloseStateStorageOnDisconnect) -
Method Summary
Modifier and TypeMethodDescriptionvoidvoidaddNimbusHost(String nimbusId, NimbusSummary nimbusSummary) voidaddPrivateWorkerKey(WorkerTokenServiceType type, String topologyId, long keyVersion, PrivateWorkerKey key) Store a new version of a private key.assignmentInfo(String stormId, Runnable callback) Get the assignment based on storm id from local backend.assignmentInfoWithVersion(String stormId, Runnable callback) assignments(Runnable callback) Get all the topologies assignments mapping stormId -> Assignment from local backend.assignmentVersion(String stormId, Runnable callback) Get backpressure topologies.blobstoreInfo(String blobKey) credentials(String stormId, Runnable callback) voiddeleteTopologyProfileRequests(String stormId, ProfileRequest profileRequest) voidneed to take executor->node+port in explicitly so that we don't run into a situation where a long dead worker with a skewed clock overrides all the timestamps.Get leader info from state store, which was written when a master gains leadership.longgetNextPrivateWorkerKeyVersion(WorkerTokenServiceType type, String topologyId) Get the next key version number that should be used for this topology id.getPrivateWorkerKey(WorkerTokenServiceType type, String topologyId, long keyVersion) Get a private key used to validate a token is correct.getTopologyProfileRequests(String stormId) getWorkerHeartbeat(String stormId, String node, Long port) getWorkerProfileRequests(String stormId, NodeInfo nodeInfo) Get a list of all topologyIds that currently have private worker keys stored, of any kind.booleanFlag to indicate if the assignments synced successfully, seeIStormClusterState.syncRemoteAssignments(Map).booleanFlag to indicate if the Pacameker is backend store.protected voidprotected voidissueMapCallback(ConcurrentHashMap<String, Runnable> callbackConcurrentHashMap, String key) nimbuses()remoteAssignmentInfo(String stormId, Runnable callback) Get the assignment based on storm id from remote state store, eg: ZK.voidremoveAllPrivateWorkerKeys(String topologyId) Remove all of the worker keys for a given topology.voidremoveBackpressure(String stormId) Remove backpressure.voidremoveBlobstoreKey(String blobKey) voidremoveExpiredPrivateWorkerKeys(String topologyId) Remove all keys for the given topology that have expired.voidremoveKeyVersion(String blobKey) voidremoveStorm(String stormId) voidremoveStormBase(String stormId) voidremoveWorkerBackpressure(String stormId, String node, Long port) Remove worker backpressure.voidremoveWorkerHeartbeat(String stormId, String node, Long port) voidvoidsetAssignment(String stormId, Assignment info, Map<String, Object> topoConf) voidMark the assignments as synced successfully, seeIStormClusterState.isAssignmentsBackendSynchronized().voidsetCredentials(String stormId, Credentials creds, Map<String, Object> topoConf) voidvoidsetupBackpressure(String stormId, Map<String, Object> topoConf) Setup backpressure.voidsetupBlob(String key, NimbusInfo nimbusInfo, Integer versionInfo) voidsetupErrors(String stormId, Map<String, Object> topoConf) voidsetupHeatbeats(String stormId, Map<String, Object> topoConf) voidsetWorkerProfileRequest(String stormId, ProfileRequest profileRequest) Get a storm base for a topology.Get storm id from passed name, null if the name doesn't exist on cluster.voidsupervisorHeartbeat(String supervisorId, SupervisorInfo info) supervisorInfo(String supervisorId) supervisors(Runnable callback) voidsyncRemoteAssignments(Map<String, byte[]> remote) Sync the remote state store assignments to local backend, used when master gains leadership, seeorg.apache.storm.nimbus.LeaderListenerCallback.voidsyncRemoteIds(Map<String, String> remote) Sync all the active storm ids of the cluster, used now when master gains leadership.voidteardownHeartbeats(String stormId) voidteardownTopologyErrors(String stormId) booleantopologyBackpressure(String stormId, long timeoutMs, Runnable callback) Check whether a topology is in throttle-on status or not: if the backpresure/storm-id dir is not empty, this topology has throttle-on, otherwise throttle-off.topologyLogConfig(String stormId, Runnable cb) voidupdateStorm(String stormId, StormBase newElems) To update this function due to APersistentMap/APersistentSet is clojure's structure.voidworkerHeartbeat(String stormId, String node, Long port, ClusterWorkerHeartbeat info) Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, waitMethods inherited from interface org.apache.storm.cluster.IStormClusterState
allSupervisorInfo, allSupervisorInfo, getTopoId, topologyBases
-
Constructor Details
-
StormClusterStateImpl
public StormClusterStateImpl(IStateStorage stateStorage, ILocalAssignmentsBackend assignmentsassignmentsBackend, ClusterStateContext context, boolean shouldCloseStateStorageOnDisconnect) throws Exception - Throws:
Exception
-
-
Method Details
-
issueCallback
-
issueMapCallback
protected void issueMapCallback(ConcurrentHashMap<String, Runnable> callbackConcurrentHashMap, String key) -
assignments
- Specified by:
assignmentsin interfaceIStormClusterState
-
assignmentInfo
Description copied from interface:IStormClusterStateGet the assignment based on storm id from local backend.- Specified by:
assignmentInfoin interfaceIStormClusterState- Parameters:
stormId- topology idcallback- callback function- Returns:
Assignment
-
remoteAssignmentInfo
Description copied from interface:IStormClusterStateGet the assignment based on storm id from remote state store, eg: ZK.- Specified by:
remoteAssignmentInfoin interfaceIStormClusterState- Parameters:
stormId- topology idcallback- callback function- Returns:
Assignment
-
assignmentsInfo
Description copied from interface:IStormClusterStateGet all the topologies assignments mapping stormId -> Assignment from local backend.- Specified by:
assignmentsInfoin interfaceIStormClusterState- Returns:
- stormId -> Assignment mapping
-
syncRemoteAssignments
Description copied from interface:IStormClusterStateSync the remote state store assignments to local backend, used when master gains leadership, seeorg.apache.storm.nimbus.LeaderListenerCallback.- Specified by:
syncRemoteAssignmentsin interfaceIStormClusterState- Parameters:
remote- assigned assignments for a specificIStormClusterStateinstance, usually a supervisor/node.
-
isAssignmentsBackendSynchronized
public boolean isAssignmentsBackendSynchronized()Description copied from interface:IStormClusterStateFlag to indicate if the assignments synced successfully, seeIStormClusterState.syncRemoteAssignments(Map).- Specified by:
isAssignmentsBackendSynchronizedin interfaceIStormClusterState- Returns:
- true if is synced successfully
-
isPacemakerStateStore
public boolean isPacemakerStateStore()Description copied from interface:IStormClusterStateFlag to indicate if the Pacameker is backend store.- Specified by:
isPacemakerStateStorein interfaceIStormClusterState- Returns:
- true if Pacemaker is being used as StateStore
-
setAssignmentsBackendSynchronized
public void setAssignmentsBackendSynchronized()Description copied from interface:IStormClusterStateMark the assignments as synced successfully, seeIStormClusterState.isAssignmentsBackendSynchronized().- Specified by:
setAssignmentsBackendSynchronizedin interfaceIStormClusterState
-
assignmentInfoWithVersion
- Specified by:
assignmentInfoWithVersionin interfaceIStormClusterState
-
assignmentVersion
- Specified by:
assignmentVersionin interfaceIStormClusterState- Throws:
Exception
-
blobstoreInfo
- Specified by:
blobstoreInfoin interfaceIStormClusterState
-
nimbuses
- Specified by:
nimbusesin interfaceIStormClusterState
-
addNimbusHost
- Specified by:
addNimbusHostin interfaceIStormClusterState
-
activeStorms
- Specified by:
activeStormsin interfaceIStormClusterState
-
stormBase
Description copied from interface:IStormClusterStateGet a storm base for a topology.- Specified by:
stormBasein interfaceIStormClusterState- Parameters:
stormId- the id of the topologycallback- something to call if the data changes (best effort)- Returns:
- the StormBase or null if it is not alive.
-
stormId
Description copied from interface:IStormClusterStateGet storm id from passed name, null if the name doesn't exist on cluster.- Specified by:
stormIdin interfaceIStormClusterState- Parameters:
stormName- storm name- Returns:
- storm id
-
syncRemoteIds
Description copied from interface:IStormClusterStateSync all the active storm ids of the cluster, used now when master gains leadership.- Specified by:
syncRemoteIdsin interfaceIStormClusterState- Parameters:
remote- stormName -> stormId mapping
-
getWorkerHeartbeat
- Specified by:
getWorkerHeartbeatin interfaceIStormClusterState
-
getWorkerProfileRequests
- Specified by:
getWorkerProfileRequestsin interfaceIStormClusterState
-
getTopologyProfileRequests
- Specified by:
getTopologyProfileRequestsin interfaceIStormClusterState
-
setWorkerProfileRequest
- Specified by:
setWorkerProfileRequestin interfaceIStormClusterState
-
deleteTopologyProfileRequests
- Specified by:
deleteTopologyProfileRequestsin interfaceIStormClusterState
-
executorBeats
public Map<ExecutorInfo,ExecutorBeat> executorBeats(String stormId, Map<List<Long>, NodeInfo> executorNodePort) need to take executor->node+port in explicitly so that we don't run into a situation where a long dead worker with a skewed clock overrides all the timestamps. By only checking heartbeats with an assigned node+port, and only reading executors from that heartbeat that are actually assigned, we avoid situations like that.- Specified by:
executorBeatsin interfaceIStormClusterState- Parameters:
stormId- topology idexecutorNodePort- executor id -> node + port- Returns:
- mapping of executorInfo -> executor beat
-
supervisors
- Specified by:
supervisorsin interfaceIStormClusterState
-
supervisorInfo
- Specified by:
supervisorInfoin interfaceIStormClusterState
-
setupHeatbeats
- Specified by:
setupHeatbeatsin interfaceIStormClusterState
-
teardownHeartbeats
- Specified by:
teardownHeartbeatsin interfaceIStormClusterState
-
teardownTopologyErrors
- Specified by:
teardownTopologyErrorsin interfaceIStormClusterState
-
getLeader
Description copied from interface:IStormClusterStateGet leader info from state store, which was written when a master gains leadership.Caution: it can not be used for fencing and is only for informational purposes because we use ZK as our backend now, which could have a overdue info of nodes.
- Specified by:
getLeaderin interfaceIStormClusterState- Parameters:
callback- callback func- Returns:
NimbusInfo
-
backpressureTopologies
Description copied from interface:IStormClusterStateGet backpressure topologies. Note: In Storm 2.0. Retained for enabling transition from 1.x. Will be removed soon.- Specified by:
backpressureTopologiesin interfaceIStormClusterState
-
heartbeatStorms
- Specified by:
heartbeatStormsin interfaceIStormClusterState
-
errorTopologies
- Specified by:
errorTopologiesin interfaceIStormClusterState
-
setTopologyLogConfig
- Specified by:
setTopologyLogConfigin interfaceIStormClusterState
-
topologyLogConfig
- Specified by:
topologyLogConfigin interfaceIStormClusterState
-
workerHeartbeat
- Specified by:
workerHeartbeatin interfaceIStormClusterState
-
removeWorkerHeartbeat
- Specified by:
removeWorkerHeartbeatin interfaceIStormClusterState
-
supervisorHeartbeat
- Specified by:
supervisorHeartbeatin interfaceIStormClusterState
-
topologyBackpressure
Check whether a topology is in throttle-on status or not: if the backpresure/storm-id dir is not empty, this topology has throttle-on, otherwise throttle-off. But if the backpresure/storm-id dir is not empty and has not been updated for more than timeoutMs, we treat it as throttle-off. This will prevent the spouts from getting stuck indefinitely if something wrong happens.- Specified by:
topologyBackpressurein interfaceIStormClusterState- Parameters:
stormId- The topology IdtimeoutMs- How long until the backpressure znode is invalid.callback- The callback function- Returns:
- True is backpresure/storm-id dir is not empty and at least one of the backpressure znodes has not timed out; false otherwise.
-
setupBackpressure
Description copied from interface:IStormClusterStateSetup backpressure. Note: In Storm 2.0. Retained for enabling transition from 1.x. Will be removed soon.- Specified by:
setupBackpressurein interfaceIStormClusterState
-
removeBackpressure
Description copied from interface:IStormClusterStateRemove backpressure. Note: In Storm 2.0. Retained for enabling transition from 1.x. Will be removed soon.- Specified by:
removeBackpressurein interfaceIStormClusterState
-
removeWorkerBackpressure
Description copied from interface:IStormClusterStateRemove worker backpressure. Note: In Storm 2.0. Retained for enabling transition from 1.x. Will be removed soon.- Specified by:
removeWorkerBackpressurein interfaceIStormClusterState
-
activateStorm
- Specified by:
activateStormin interfaceIStormClusterState
-
updateStorm
To update this function due to APersistentMap/APersistentSet is clojure's structure.- Specified by:
updateStormin interfaceIStormClusterState
-
removeStormBase
- Specified by:
removeStormBasein interfaceIStormClusterState
-
setAssignment
- Specified by:
setAssignmentin interfaceIStormClusterState
-
setupBlob
- Specified by:
setupBlobin interfaceIStormClusterState
-
activeKeys
- Specified by:
activeKeysin interfaceIStormClusterState
-
blobstore
- Specified by:
blobstorein interfaceIStormClusterState
-
removeStorm
- Specified by:
removeStormin interfaceIStormClusterState
-
removeBlobstoreKey
- Specified by:
removeBlobstoreKeyin interfaceIStormClusterState
-
removeKeyVersion
- Specified by:
removeKeyVersionin interfaceIStormClusterState
-
setupErrors
- Specified by:
setupErrorsin interfaceIStormClusterState
-
reportError
public void reportError(String stormId, String componentId, String node, Long port, Throwable error) - Specified by:
reportErrorin interfaceIStormClusterState
-
errors
- Specified by:
errorsin interfaceIStormClusterState
-
lastError
- Specified by:
lastErrorin interfaceIStormClusterState
-
setCredentials
- Specified by:
setCredentialsin interfaceIStormClusterState
-
credentials
- Specified by:
credentialsin interfaceIStormClusterState
-
disconnect
public void disconnect()- Specified by:
disconnectin interfaceIStormClusterState
-
getPrivateWorkerKey
public PrivateWorkerKey getPrivateWorkerKey(WorkerTokenServiceType type, String topologyId, long keyVersion) Description copied from interface:IStormClusterStateGet a private key used to validate a token is correct. This is expected to be called from a privileged daemon, and the ACLs should be set up to only allow nimbus and these privileged daemons access to these private keys.- Specified by:
getPrivateWorkerKeyin interfaceIStormClusterState- Parameters:
type- the type of service the key is for.topologyId- the topology id the key is for.keyVersion- the version of the key this is for.- Returns:
- the private key or null if it could not be found.
-
addPrivateWorkerKey
public void addPrivateWorkerKey(WorkerTokenServiceType type, String topologyId, long keyVersion, PrivateWorkerKey key) Description copied from interface:IStormClusterStateStore a new version of a private key. This is expected to only ever be called from nimbus. All ACLs however need to be setup to allow the given services access to the stored information.- Specified by:
addPrivateWorkerKeyin interfaceIStormClusterState- Parameters:
type- the type of service this key is for.topologyId- the topology this key is forkeyVersion- the version of the key this is for.key- the key to store.
-
getNextPrivateWorkerKeyVersion
Description copied from interface:IStormClusterStateGet the next key version number that should be used for this topology id. This is expected to only ever be called from nimbus, but it is acceptable if the ACLs are setup so that it can work from a privileged daemon for the given service.- Specified by:
getNextPrivateWorkerKeyVersionin interfaceIStormClusterState- Parameters:
type- the type of service this is for.topologyId- the topology id this is for.- Returns:
- the next version number. It should be 0 for a new topology id/service combination.
-
removeExpiredPrivateWorkerKeys
Description copied from interface:IStormClusterStateRemove all keys for the given topology that have expired. The number of keys should be small enough that doing an exhaustive scan of them all is acceptable as there is no guarantee that expiration time and version number are related. This should be for all service types. This is expected to only ever be called from nimbus and some ACLs may be setup so being called from other daemons will cause it to fail.- Specified by:
removeExpiredPrivateWorkerKeysin interfaceIStormClusterState- Parameters:
topologyId- the id of the topology to scan.
-
removeAllPrivateWorkerKeys
Description copied from interface:IStormClusterStateRemove all of the worker keys for a given topology. Used to clean up after a topology finishes. This is expected to only ever be called from nimbus and ideally should only ever work from nimbus.- Specified by:
removeAllPrivateWorkerKeysin interfaceIStormClusterState- Parameters:
topologyId- the topology to clean up after.
-
idsOfTopologiesWithPrivateWorkerKeys
Description copied from interface:IStormClusterStateGet a list of all topologyIds that currently have private worker keys stored, of any kind. This is expected to only ever be called from nimbus.- Specified by:
idsOfTopologiesWithPrivateWorkerKeysin interfaceIStormClusterState- Returns:
- the list of topology ids with any kind of private worker key stored.
-