Package org.apache.solr.cluster.placement.plugins

Class AffinityPlacementConfig

java.lang.Object
org.apache.solr.cluster.placement.plugins.AffinityPlacementConfig
All Implemented Interfaces:
PlacementPluginConfig, org.apache.solr.common.MapSerializable, org.apache.solr.common.MapWriter, org.apache.solr.common.NavigableObject, org.apache.solr.common.util.ReflectMapWriter, org.noggit.JSONWriter.Writable
public class AffinityPlacementConfig extends Object implements PlacementPluginConfig
Configuration bean for AffinityPlacementFactory.

  • Nested Class Summary

    Nested classes/interfaces inherited from interface org.apache.solr.common.MapWriter

    org.apache.solr.common.MapWriter.EntryWriter

  • Field Summary

    Fields
    Modifier and Type
    Field
    Description
    static final String
    AVAILABILITY_ZONE_SYSPROP
    Name of the system property on a node indicating which (public cloud) Availability Zone that node is in.
    Map<String,String>
    collectionNodeType
    This property defines an additional constraint that the collection must be placed only on the nodes of the correct "node type".
    static final AffinityPlacementConfig
    DEFAULT
     
    static final long
    DEFAULT_MINIMAL_FREE_DISK_GB
     
    static final long
    DEFAULT_PRIORITIZED_FREE_DISK_GB
     
    Integer
    maxReplicasPerShardInDomain
    Determines the maximum number of replicas of a particular type of a particular shard that can be placed within a single domain (as defined by the @link #SPREAD_DOMAIN_SYSPROP} System property.
    long
    minimalFreeDiskGB
    If a node has strictly less GB of free disk than this value, the node is excluded from assignment decisions.
    static final String
    NODE_TYPE_SYSPROP
    Name of the system property on a node indicating the arbitrary "node type" (for example, a node more suitable for the indexing work load could be labeled as node_type: indexing).
    long
    prioritizedFreeDiskGB
    Replica allocation will assign replicas to nodes with at least this number of GB of free disk space regardless of the number of cores on these nodes rather than assigning replicas to nodes with less than this amount of free disk space if that's an option (if that's not an option, replicas can still be assigned to nodes with less than this amount of free space).
    static final String
    REPLICA_TYPE_SYSPROP
    Name of the system property on a node indicating the type of replicas allowed on that node.
    static final String
    SPREAD_DOMAIN_SYSPROP
    Name of the system property on a node indicating the spread domain group.
    Boolean
    spreadAcrossDomains
    When this property is set to true, Solr will try to place replicas for the same shard in nodes that have different value for the SPREAD_DOMAIN_SYSPROP System property.
    static final String
    UNDEFINED_AVAILABILITY_ZONE
    This is the "AZ" name for nodes that do not define an AZ.
    Map<String,String>
    withCollection
    This property defines an additional constraint that primary collections (keys) should be located on the same nodes as the secondary collections (values).
    Map<String,String>
    withCollectionShards
    Same as withCollection but ensures shard to shard correspondence.

    Fields inherited from interface org.apache.solr.common.MapWriter

    EMPTY

  • Constructor Summary

    Constructors
    Constructor
    Description
    AffinityPlacementConfig()
    Zero-arguments public constructor required for deserialization - don't use.
    AffinityPlacementConfig(long minimalFreeDiskGB, long prioritizedFreeDiskGB)
    Configuration for the AffinityPlacementFactory.
    AffinityPlacementConfig(long minimalFreeDiskGB, long prioritizedFreeDiskGB, Map<String,String> withCollection, Map<String,String> collectionNodeType)
     
    AffinityPlacementConfig(long minimalFreeDiskGB, long prioritizedFreeDiskGB, Map<String,String> withCollection, Map<String,String> withCollectionShards, Map<String,String> collectionNodeType)
    Configuration for the AffinityPlacementFactory.

  • Method Summary

    Modifier and Type
    Method
    Description
    void
    validate()
     

    Methods inherited from class java.lang.Object

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

    Methods inherited from interface org.apache.solr.common.MapWriter

    jsonStr, toMap, write

    Methods inherited from interface org.apache.solr.common.NavigableObject

    _forEachEntry, _forEachEntry, _forEachEntry, _get, _get, _getStr, _getStr, _size

    Methods inherited from interface org.apache.solr.common.util.ReflectMapWriter

    writeMap

  • Field Details

    • DEFAULT_MINIMAL_FREE_DISK_GB

      public static final long DEFAULT_MINIMAL_FREE_DISK_GB
      See Also:

    • DEFAULT_PRIORITIZED_FREE_DISK_GB

      public static final long DEFAULT_PRIORITIZED_FREE_DISK_GB
      See Also:

    • DEFAULT

      public static final AffinityPlacementConfig DEFAULT

    • AVAILABILITY_ZONE_SYSPROP

      public static final String AVAILABILITY_ZONE_SYSPROP
      Name of the system property on a node indicating which (public cloud) Availability Zone that node is in. The value is any string, different strings denote different availability zones.

      Nodes on which this system property is not defined are considered being in the same Availability Zone UNDEFINED_AVAILABILITY_ZONE (hopefully the value of this constant is not the name of a real Availability Zone :).

      See Also:

    • REPLICA_TYPE_SYSPROP

      public static final String REPLICA_TYPE_SYSPROP
      Name of the system property on a node indicating the type of replicas allowed on that node. The value of that system property is a comma separated list or a single string of value names of Replica.ReplicaType (case insensitive). If that property is not defined, that node is considered accepting all replica types (i.e. undefined is equivalent to "NRT,Pull,tlog").
      See Also:

    • NODE_TYPE_SYSPROP

      public static final String NODE_TYPE_SYSPROP
      Name of the system property on a node indicating the arbitrary "node type" (for example, a node more suitable for the indexing work load could be labeled as node_type: indexing). The value of this system property is a comma-separated list or a single label (labels must not contain commas), which represent a logical OR for the purpose of placement.
      See Also:

    • SPREAD_DOMAIN_SYSPROP

      public static final String SPREAD_DOMAIN_SYSPROP
      Name of the system property on a node indicating the spread domain group. This is used (if spreadAcrossDomains is set to true) to indicate this placement plugin that replicas for a particular shard should spread across nodes that have different values for this system property.
      See Also:

    • UNDEFINED_AVAILABILITY_ZONE

      public static final String UNDEFINED_AVAILABILITY_ZONE
      This is the "AZ" name for nodes that do not define an AZ. Should not match a real AZ name (I think we're safe)
      See Also:

    • minimalFreeDiskGB

      public long minimalFreeDiskGB
      If a node has strictly less GB of free disk than this value, the node is excluded from assignment decisions. Set to 0 or less to disable.

    • prioritizedFreeDiskGB

      public long prioritizedFreeDiskGB
      Replica allocation will assign replicas to nodes with at least this number of GB of free disk space regardless of the number of cores on these nodes rather than assigning replicas to nodes with less than this amount of free disk space if that's an option (if that's not an option, replicas can still be assigned to nodes with less than this amount of free space).

    • withCollection

      public Map<String,String> withCollection
      This property defines an additional constraint that primary collections (keys) should be located on the same nodes as the secondary collections (values). The plugin will assume that the secondary collection replicas are already in place and ignore candidate nodes where they are not already present.

    • collectionNodeType

      public Map<String,String> collectionNodeType
      This property defines an additional constraint that the collection must be placed only on the nodes of the correct "node type". The nodes can specify what type they are (one or several types, using a comma-separated list) by defining the NODE_TYPE_SYSPROP system property. Similarly, the plugin can be configured to specify that a collection (key in the map) must be placed on one or more node type (value in the map, using comma-separated list of acceptable node types).

    • withCollectionShards

      public Map<String,String> withCollectionShards
      Same as withCollection but ensures shard to shard correspondence. should be disjoint with withCollection.

    • spreadAcrossDomains

      public Boolean spreadAcrossDomains
      When this property is set to true, Solr will try to place replicas for the same shard in nodes that have different value for the SPREAD_DOMAIN_SYSPROP System property. If more replicas exist (or are being placed) than the number of different values for SPREAD_DOMAIN_SYSPROP System property in nodes in the cluster, Solr will attempt to distribute the placement of the replicas evenly across the domains but will fail the placement if more than maxReplicasPerShardInDomain are placed within a single domain. Note that the domain groups are evaluated within a particular AZ (i.e. Solr will not consider the placement of replicas in AZ1 when selecting candidate nodes for replicas in AZ2). Example usages for this config are:
      • Rack diversity: You want replicas in different AZs but also, within the AZ you want them in different racks
      • Host diversity: You are running multiple Solr instances in the same host physical host. You want replicas in different AZs but also, within an AZ you want replicas for the same shard to go in nodes that run in different hosts

    • maxReplicasPerShardInDomain

      public Integer maxReplicasPerShardInDomain
      Determines the maximum number of replicas of a particular type of a particular shard that can be placed within a single domain (as defined by the @link #SPREAD_DOMAIN_SYSPROP} System property.

  • Constructor Details

    • AffinityPlacementConfig

      public AffinityPlacementConfig()
      Zero-arguments public constructor required for deserialization - don't use.

    • AffinityPlacementConfig

      public AffinityPlacementConfig(long minimalFreeDiskGB, long prioritizedFreeDiskGB)
      Configuration for the AffinityPlacementFactory.
      Parameters:
      minimalFreeDiskGB - minimal free disk GB.
      prioritizedFreeDiskGB - prioritized free disk GB.

    • AffinityPlacementConfig

      public AffinityPlacementConfig(long minimalFreeDiskGB, long prioritizedFreeDiskGB, Map<String,String> withCollection, Map<String,String> withCollectionShards, Map<String,String> collectionNodeType)
      Configuration for the AffinityPlacementFactory.
      Parameters:
      minimalFreeDiskGB - minimal free disk GB.
      prioritizedFreeDiskGB - prioritized free disk GB.
      withCollection - configuration of co-located collections: keys are primary collection names and values are secondary collection names.
      collectionNodeType - configuration of reequired node types per collection. Keys are collection names and values are comma-separated lists of required node types.

    • AffinityPlacementConfig

      public AffinityPlacementConfig(long minimalFreeDiskGB, long prioritizedFreeDiskGB, Map<String,String> withCollection, Map<String,String> collectionNodeType)

  • Method Details

    • validate

      public void validate()