Package com.illumon.iris.controller

Class PersistentQueryController

java.lang.Object

com.fishlib.util.process.OnetimeShutdownTask

com.illumon.iris.controller.PersistentQueryController

All Implemented Interfaces:

com.fishlib.util.process.ShutdownManager.Task

public final class PersistentQueryController
extends com.fishlib.util.process.OnetimeShutdownTask

Stand-alone controller for Iris.

Hosts a configuration PHT, a state PDHT, monitors server status, and dispatches stop/start commands.

Thread-safety notes:

• All changes to IrisQueryInfos are protected by the lock on the associated PDHT's map.
• All PersistentQueryConfiguration instances that are published (in the associated PHT or PersistentQueryInfo) are guaranteed not to be further mutated. Hence, they can be read safely at all times.
• All changes to a PersistentQueryConfiguration (by serial, not by instance, since published instances are immutable) are protected by a serial-specific lock. Query starts/stops are also protected, since they must always use the most up-to-date version of the configuration.

Configuration properties

• PersistentQueryController.port - The port for communication
• PersistentQueryController.useLocalGit (optional) - When set to true use the local filesystem as a repository
• PersistentQueryController.gitGcEnabled (optional) - When set to true a git gc will be periodically performed
• PersistentQueryController.statusCheckTimes - When an AnomalyClient is used, defines the times to check in with status
• PersistentQueryController.failureWindowMillis - How long failures are queued
• PersistentQueryController.failureNumQueriesThreshold - How many failures are required before notifying the AnomalyClient
• PersistentQueryController.failureResendMillis - The minimum gap between failure messages
• PersistentQueryController.commitCheckpointImmediate (optional) - Immediately checkpoint the PDHT state on startup
• RemoteProcessingRequestProfile.defaultProfile - The default JVM args profile to use
• RemoteProcessingRequestProfile.noneProfile - The JVM profile to use when "None" is selected

Field Summary

Fields

Modifier and Type	Field	Description
static final String	CONTROLLER_ELECTION_CAMPAIGN_KEY
static final int	CONTROLLER_PRESENCE_KV_TTL_SECONDS
static final double	DEFAULT_MAX_HEAP_SIZE_GB
static final long	ELECTION_OBSERVE_TIMEOUT_MILLIS
static final long	ELECTION_RPC_TIMEOUT_MILLIS
static final int	ETCD_ELECTION_LEASE_KEEPALIVE_TIMEOUT_MS	How long will we wait for our etcd election lease to be granted or renewed, in milliseconds before determining that we had a failure.
static final int	ETCD_ELECTION_LEASE_RENEW_RETRY_DELAY_SECONDS	What is the maximum frequency which we will try to renew an election lease.
static final int	ETCD_ELECTION_LEASE_RENEW_SECONDS	How often, in seconds, do we renew our etcd lease for leader election.
static final int	ETCD_ELECTION_LEASE_TTL_SECONDS	How long does our etcd lease for leader election last, in seconds? If we do not renew the lease for this many seconds, then etcd will automatically revoke it, and we will lose our leadership and our registration in dh-resolver will be removed.
static final int	ETCD_STATE_KVS_LOAD_BATCH_SIZE
static final int	ETCD_STATE_KVS_LOAD_TIMEOUT_SECONDS	How long in seconds the controller wait to load existing worker state.
static final int	ETCD_STATE_LEASE_KEEPALIVE_TIMEOUT_MS	How long will we wait for our etcd PQ State lease to be granted or renewed, in milliseconds before determining that we had a failure.
static final int	ETCD_STATE_LEASE_RENEW_RETRY_DELAY_SECONDS	What is the maximum frequency which we will try to renew a PQ State lease.
static final int	ETCD_STATE_LEASE_RENEW_SECONDS	How often, in seconds, do we renew our etcd lease for PQ State KVs.
static final int	ETCD_STATE_LEASE_TTL_SECONDS	How long does our etcd lease for PQ state last, in seconds? If we do not renew the lease for this many seconds, then etcd will automatically revoke it, and our presence KV will be removed.
static final int	MAX_DESYNC_RETRIES
static final int	MAX_POSSIBLE_HEAP_GB
static final String	RESET_ON_STARTUP

Constructor Summary

Constructors

Constructor

Description

PersistentQueryController(String processNameForLogging, String processInfoId, AuditEventLogger auditEventLogger, PersistentQueryStateLogger stateLogLogger, PersistentQueryConfigurationLogger configurationLogLogger, ResourceUtilizationLogger resourceUtilizationLogger, QueryUserAssignmentLogger queryUserAssignmentLogger, io.deephaven.enterprise.niowrapper.sched.Scheduler scheduler, io.deephaven.enterprise.comm.ConnectionMonitor.Params connectionMonitorParams, io.deephaven.enterprise.ProcessStatusReporter chatBotClient, ScriptPathLoaderManager scriptPathLoaderManager, String errorEmailRecipientAddress, IrisGroupProvider irisGroupProvider, ScheduledExecutorService executorService, AuthenticationClient authClient, Function<String,PublicPrivateKeyEncryption> encryptionFactory, ControllerDispatcherHandler.ConnectionFactory dispatcherConnectionFactory, io.deephaven.shadow.jetcd.io.etcd.jetcd.Client etcdClient, io.deephaven.enterprise.etcdclient.presence.PresenceKvLeaseHandler pqStateLeaseHandler, byte[] uuid, io.deephaven.enterprise.etcdclient.lease.LeaseHandler electionLeaseHandler, io.deephaven.enterprise.etcdclient.election.ElectionCandidate electionCandidate)

Method Summary

Modifier and Type	Method	Description
AuthenticationClient	getAuthClient()
PersistentQueryConfigurationLogger	getConfigurationLogLogger()
IrisGroupProvider	getGroupProvider()
QueryHandleContext.PresenceManager	getPresenceManager()
String	getSelectionProviderStatus(UserContext userContext)	Query the controller's active IServerSelectionProvider for status.
Map<String,WorkerKind>	getWorkerKindMap()
static void	main(String... args)
String	selectionProviderCommand(UserContext userContext, String command)	Provide a String command to the controller's active IServerSelectionProvider.
void	shutdown()
void	start(io.deephaven.shadow.core.io.grpc.ServerBuilder<?> serverBuilder, Supplier<PqStorage> pqStorageFactory, boolean loadStateFromPresenceKvs)

Methods inherited from class com.fishlib.util.process.OnetimeShutdownTask

adapt, awaitShutdown, awaitShutdown, invoke, isShutdown

Methods inherited from class java.lang.Object

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

Field Details

RESET_ON_STARTUP

public static final String RESET_ON_STARTUP

See Also:

• Constant Field Values

MAX_POSSIBLE_HEAP_GB

public static final int MAX_POSSIBLE_HEAP_GB

See Also:

• Constant Field Values

DEFAULT_MAX_HEAP_SIZE_GB

public static final double DEFAULT_MAX_HEAP_SIZE_GB

MAX_DESYNC_RETRIES

public static final int MAX_DESYNC_RETRIES

CONTROLLER_PRESENCE_KV_TTL_SECONDS

public static final int CONTROLLER_PRESENCE_KV_TTL_SECONDS

ETCD_STATE_LEASE_TTL_SECONDS

public static final int ETCD_STATE_LEASE_TTL_SECONDS

How long does our etcd lease for PQ state last, in seconds? If we do not renew the lease for this many seconds, then etcd will automatically revoke it, and our presence KV will be removed. Any process using our presence KV as sign of liveness will consider us dead.

ETCD_STATE_LEASE_RENEW_SECONDS

public static final int ETCD_STATE_LEASE_RENEW_SECONDS

How often, in seconds, do we renew our etcd lease for PQ State KVs. This must be less than ETCD_STATE_LEASE_TTL_SECONDS; for safety it should be no more than 50% of the TTL value.

ETCD_STATE_LEASE_RENEW_RETRY_DELAY_SECONDS

public static final int ETCD_STATE_LEASE_RENEW_RETRY_DELAY_SECONDS

What is the maximum frequency which we will try to renew a PQ State lease. If a lease renewal times out, we will try again more quickly than the standard ETCD_STATE_LEASE_RENEW_SECONDS.

ETCD_STATE_LEASE_KEEPALIVE_TIMEOUT_MS

public static final int ETCD_STATE_LEASE_KEEPALIVE_TIMEOUT_MS

How long will we wait for our etcd PQ State lease to be granted or renewed, in milliseconds before determining that we had a failure.

ETCD_ELECTION_LEASE_TTL_SECONDS

public static final int ETCD_ELECTION_LEASE_TTL_SECONDS

How long does our etcd lease for leader election last, in seconds? If we do not renew the lease for this many seconds, then etcd will automatically revoke it, and we will lose our leadership and our registration in dh-resolver will be removed.

ETCD_ELECTION_LEASE_RENEW_SECONDS

public static final int ETCD_ELECTION_LEASE_RENEW_SECONDS

How often, in seconds, do we renew our etcd lease for leader election. This must be less than ETCD_ELECTION_LEASE_TTL_SECONDS; for safety it should be no more than 50% of the TTL value.

ETCD_ELECTION_LEASE_RENEW_RETRY_DELAY_SECONDS

public static final int ETCD_ELECTION_LEASE_RENEW_RETRY_DELAY_SECONDS

What is the maximum frequency which we will try to renew an election lease. If a lease renewal times out, we will try again more quickly than the standard ETCD_ELECTION_LEASE_RENEW_SECONDS.

ETCD_ELECTION_LEASE_KEEPALIVE_TIMEOUT_MS

public static final int ETCD_ELECTION_LEASE_KEEPALIVE_TIMEOUT_MS

How long will we wait for our etcd election lease to be granted or renewed, in milliseconds before determining that we had a failure.

ETCD_STATE_KVS_LOAD_TIMEOUT_SECONDS

public static final int ETCD_STATE_KVS_LOAD_TIMEOUT_SECONDS

How long in seconds the controller wait to load existing worker state.

ETCD_STATE_KVS_LOAD_BATCH_SIZE

public static final int ETCD_STATE_KVS_LOAD_BATCH_SIZE

CONTROLLER_ELECTION_CAMPAIGN_KEY

public static final String CONTROLLER_ELECTION_CAMPAIGN_KEY

See Also:

• Constant Field Values

ELECTION_RPC_TIMEOUT_MILLIS

public static final long ELECTION_RPC_TIMEOUT_MILLIS

ELECTION_OBSERVE_TIMEOUT_MILLIS

public static final long ELECTION_OBSERVE_TIMEOUT_MILLIS

Constructor Details

PersistentQueryController

public PersistentQueryController(@NotNull
 String processNameForLogging,
 @NotNull
 String processInfoId,
 @NotNull
 AuditEventLogger auditEventLogger,
 @NotNull
 PersistentQueryStateLogger stateLogLogger,
 @NotNull
 PersistentQueryConfigurationLogger configurationLogLogger,
 @NotNull
 ResourceUtilizationLogger resourceUtilizationLogger,
 @NotNull
 QueryUserAssignmentLogger queryUserAssignmentLogger,
 @NotNull
 io.deephaven.enterprise.niowrapper.sched.Scheduler scheduler,
 @Nullable
 io.deephaven.enterprise.comm.ConnectionMonitor.Params connectionMonitorParams,
 @Nullable
 io.deephaven.enterprise.ProcessStatusReporter chatBotClient,
 @NotNull
 ScriptPathLoaderManager scriptPathLoaderManager,
 @Nullable
 String errorEmailRecipientAddress,
 @NotNull
 IrisGroupProvider irisGroupProvider,
 @NotNull
 ScheduledExecutorService executorService,
 @NotNull
 AuthenticationClient authClient,
 @NotNull
 Function<String,PublicPrivateKeyEncryption> encryptionFactory,
 @NotNull
 ControllerDispatcherHandler.ConnectionFactory dispatcherConnectionFactory,
 @NotNull
 io.deephaven.shadow.jetcd.io.etcd.jetcd.Client etcdClient,
 @NotNull
 io.deephaven.enterprise.etcdclient.presence.PresenceKvLeaseHandler pqStateLeaseHandler,
 @NotNull
 byte[] uuid,
 @NotNull
 io.deephaven.enterprise.etcdclient.lease.LeaseHandler electionLeaseHandler,
 @NotNull
 io.deephaven.enterprise.etcdclient.election.ElectionCandidate electionCandidate)
                          throws IOException,
ConfigurationVerificationException

Throws:

IOException

ConfigurationVerificationException

Method Details

start

public void start(@NotNull
 io.deephaven.shadow.core.io.grpc.ServerBuilder<?> serverBuilder,
 @NotNull
 Supplier<PqStorage> pqStorageFactory,
 boolean loadStateFromPresenceKvs)
           throws IOException,
PqStorageException

Throws:

IOException

PqStorageException

shutdown

public void shutdown()

Specified by:

shutdown in class com.fishlib.util.process.OnetimeShutdownTask

getSelectionProviderStatus

public String getSelectionProviderStatus(UserContext userContext)

Query the controller's active IServerSelectionProvider for status.

Parameters:

userContext - the user making the request

Returns:

a free-form string of server selection provider information

selectionProviderCommand

public String selectionProviderCommand(@NotNull
 UserContext userContext,
 String command)

Provide a String command to the controller's active IServerSelectionProvider.

This can be used to mutate the internal state of the selection provider. For example, the SimpleServerSelectionProvider enables the administrator to mark backend servers as administratively down to exclude them from the server selection logic (for new queries).

Parameters:

userContext - the user making the request

command - the free-form server-selection provider command (e.g. JSON)

Returns:

a free-form string of server selection provider command results

getGroupProvider

public IrisGroupProvider getGroupProvider()

getConfigurationLogLogger

public PersistentQueryConfigurationLogger getConfigurationLogLogger()

getAuthClient

@NotNull
public AuthenticationClient getAuthClient()

getPresenceManager

public QueryHandleContext.PresenceManager getPresenceManager()

getWorkerKindMap

public Map<String,WorkerKind> getWorkerKindMap()

main

public static void main(String... args)
                 throws IOException,
ClassNotFoundException,
ConfigurationVerificationException,
PqStorageException

Throws:

IOException

ClassNotFoundException

ConfigurationVerificationException

PqStorageException