Package io.debezium.connector.mysql.legacy

Class SourceInfo

  • @NotThreadSafe
final class SourceInfo
extends BaseSourceInfo
    Information about the source of information, which includes the position in the source binary log we have previously processed.

    The source partition information describes the database whose log is being consumed. Typically, the database is identified by the host address port number of the MySQL server and the name of the database. Here's a JSON-like representation of an example database: 

     {
     "server" : "production-server"
 }

    The source offset information is included in each event and captures where the connector should restart if this event's offset is the last one recorded. The offset includes the binlog filename, the position of the first event in the binlog, the number of events to skip, and the number of rows to also skip.

    Here's a JSON-like representation of an example offset: 

     {
     "server_id": 112233,
     "ts_sec": 1465937,
     "gtid": "db58b0ae-2c10-11e6-b284-0242ac110002:199",
     "file": "mysql-bin.000003",
     "pos" = 990,
     "event" = 0,
     "row": 0,
     "snapshot": true
 }

    The "gtids" field only appears in offsets produced when GTIDs are enabled. The "snapshot" field only appears in offsets produced when the connector is in the middle of a snapshot. And finally, the "ts" field contains the seconds since Unix epoch (since Jan 1, 1970) of the MySQL event; the message envelopes also have a timestamp, but that timestamp is the milliseconds since since Jan 1, 1970.

    Each change event envelope also includes the source struct that contains MySQL information about that particular event, including a mixture the fields from the partition (which is renamed in the source to make more sense), the binlog filename and position where the event can be found, and when GTIDs are enabled the GTID of the transaction in which the event occurs. Like with the offset, the "snapshot" field only appears for events produced when the connector is in the middle of a snapshot. Note that this information is likely different than the offset information, since the connector may need to restart from either just after the most recently completed transaction or the beginning of the most recently started transaction (whichever appears later in the binlog).

    Here's a JSON-like representation of the source for the metadata for an event that corresponds to the above partition and offset: 

     {
     "name": "production-server",
     "server_id": 112233,
     "ts_sec": 1465937,
     "gtid": "db58b0ae-2c10-11e6-b284-0242ac110002:199",
     "file": "mysql-bin.000003",
     "pos" = 1081,
     "row": 0,
     "snapshot": true,
     "thread" : 1,
     "db" : "inventory",
     "table" : "products"
 }
    Author:
    Randall Hauch

    • Method Detail

      • setQuery

        public void setQuery​(String query)
        Set the original SQL query.
        Parameters:
        query - the original SQL query that generated the event.

      • getQuery

        public String getQuery()
        Returns:
        the original SQL query that generated the event. NULL if no such query is associated.

      • partition

        public Map<String,​String> partition()
        Get the Kafka Connect detail about the source "partition", which describes the portion of the source that we are consuming. Since we're reading the binary log for a single database, the source partition specifies the database server.

        The resulting map is mutable for efficiency reasons (this information rarely changes), but should not be mutated.

        Returns:
        the source partition information; never null

      • setBinlogStartPoint

        public void setBinlogStartPoint​(String binlogFilename,
                                long positionOfFirstEvent)
        Set the position in the MySQL binlog where we will start reading.
        Parameters:
        binlogFilename - the name of the binary log file; may not be null
        positionOfFirstEvent - the position in the binary log file to begin processing

      • setEventPosition

        public void setEventPosition​(long positionOfCurrentEvent,
                             long eventSizeInBytes)
        Set the position within the MySQL binary log file of the current event.
        Parameters:
        positionOfCurrentEvent - the position within the binary log file of the current event
        eventSizeInBytes - the size in bytes of this event

      • offset

        public Map<String,​?> offset()
        Get the Kafka Connect detail about the source "offset", which describes the position within the source where we last have last read.
        Returns:
        a copy of the current offset; never null

      • offsetForRow

        public Map<String,​Object> offsetForRow​(int eventRowNumber,
                                             int totalNumberOfRows)
        Given the row number within a binlog event and the total number of rows in that event, compute and return the Kafka Connect offset that is be included in the produced change event describing the row.

        This method should always be called before AbstractSourceInfo.struct().

        Parameters:
        eventRowNumber - the 0-based row number within the event for which the offset is to be produced
        totalNumberOfRows - the total number of rows within the event being processed
        Returns:
        a copy of the current offset; never null
        See Also:
        AbstractSourceInfo.struct()

      • changeEventCompleted

        public void changeEventCompleted()

      • offsetUsingPosition

        private Map<String,​Object> offsetUsingPosition​(long rowsToSkip)

      • databaseEvent

        public void databaseEvent​(String databaseName)

      • tableEvent

        public void tableEvent​(Set<TableId> tableIds)

      • tableEvent

        public void tableEvent​(TableId tableId)

      • isSnapshotInEffect

        public boolean isSnapshotInEffect()
        Determine whether a snapshot is currently in effect.
        Returns:
        true if a snapshot is in effect, or false otherwise

      • startNextTransaction

        public void startNextTransaction()

      • completeEvent

        public void completeEvent()
        Capture that we're starting a new event.

      • eventsToSkipUponRestart

        public long eventsToSkipUponRestart()
        Get the number of events after the last transaction BEGIN that we've already processed.
        Returns:
        the number of events in the transaction that have been processed completely
        See Also:
        completeEvent(), startNextTransaction()

      • commitTransaction

        public void commitTransaction()

      • startGtid

        public void startGtid​(String gtid,
                      String gtidSet)
        Record that a new GTID transaction has been started and has been included in the set of GTIDs known to the MySQL server.
        Parameters:
        gtid - the string representation of a specific GTID that has been begun; may not be null
        gtidSet - the string representation of GTID set that includes the newly begun GTID; may not be null

      • setCompletedGtidSet

        public void setCompletedGtidSet​(String gtidSet)
        Set the GTID set that captures all of the GTID transactions that have been completely processed.
        Parameters:
        gtidSet - the string representation of the GTID set; may not be null, but may be an empty string if no GTIDs have been previously processed

      • setBinlogServerId

        public void setBinlogServerId​(long serverId)
        Set the server ID as found within the MySQL binary log file.
        Parameters:
        serverId - the server ID found within the binary log file

      • setBinlogTimestampSeconds

        public void setBinlogTimestampSeconds​(long timestampInSeconds)
        Set the number of seconds since Unix epoch (January 1, 1970) as found within the MySQL binary log file. Note that the value in the binlog events is in seconds, but the library we use returns the value in milliseconds (with only second precision and therefore all fractions of a second are zero). We capture this as seconds since that is the precision that MySQL uses.
        Parameters:
        timestampInSeconds - the timestamp in seconds found within the binary log file

      • setSourceTime

        public void setSourceTime​(Instant timestamp)

      • setBinlogThread

        public void setBinlogThread​(long threadId)
        Set the identifier of the MySQL thread that generated the most recent event.
        Parameters:
        threadId - the thread identifier; may be negative if not known

      • startSnapshot

        public void startSnapshot()
        Denote that a snapshot is being (or has been) started.

      • markLastSnapshot

        public void markLastSnapshot​(Configuration config)
        Denote that a snapshot will be complete after one last record.

      • completeSnapshot

        public void completeSnapshot()
        Denote that a snapshot has completed successfully.

      • setFilterDataFromConfig

        public void setFilterDataFromConfig​(Configuration config)
        Set the filter data for the offset from the given Configuration
        Parameters:
        config - the configuration

      • maybeSetFilterDataFromConfig

        public void maybeSetFilterDataFromConfig​(Configuration config)
        Set filter data from config if and only if parallel snapshotting of new tables is turned on
        Parameters:
        config - the configuration.

      • hasFilterInfo

        public boolean hasFilterInfo()
        Returns:
        true if this offset has filter info, false otherwise.

      • getDatabaseIncludeList

        public String getDatabaseIncludeList()

      • getDatabaseExcludeList

        public String getDatabaseExcludeList()

      • getTableIncludeList

        public String getTableIncludeList()

      • getTableExcludeList

        public String getTableExcludeList()

      • setOffset

        public void setOffset​(Map<String,​?> sourceOffset)
        Set the source offset, as read from Kafka Connect. This method does nothing if the supplied map is null.
        Parameters:
        sourceOffset - the previously-recorded Kafka Connect source offset
        Throws:
        org.apache.kafka.connect.errors.ConnectException - if any offset parameter values are missing, invalid, or of the wrong type

      • offsetsHaveFilterInfo

        public static boolean offsetsHaveFilterInfo​(Map<String,​?> sourceOffset)

      • longOffsetValue

        private long longOffsetValue​(Map<String,​?> values,
                             String key)

      • booleanOffsetValue

        private boolean booleanOffsetValue​(Map<String,​?> values,
                                   String key)

      • gtidSet

        public String gtidSet()
        Get the string representation of the GTID range for the MySQL binary log file.
        Returns:
        the string representation of the binlog GTID ranges; may be null

      • binlogFilename

        public String binlogFilename()
        Get the name of the MySQL binary log file that has last been processed.
        Returns:
        the name of the binary log file; null if it has not been set

      • binlogPosition

        public long binlogPosition()
        Get the position within the MySQL binary log file of the next event to be processed.
        Returns:
        the position within the binary log file; null if it has not been set

      • restartBinlogPosition

        protected long restartBinlogPosition()
        Get the position within the MySQL binary log file of the most recently processed event.
        Returns:
        the position within the binary log file; null if it has not been set

      • rowsToSkipUponRestart

        public int rowsToSkipUponRestart()
        Get the number of rows beyond the last completely processed event to be skipped upon restart.
        Returns:
        the number of rows to be skipped

      • getServerId

        long getServerId()

      • getThreadId

        long getThreadId()

      • table

        String table()
        Returns a string representation of the table(s) affected by the current event. Will only represent more than a single table for events in the user-facing schema history topic for certain types of DDL. Will be null for DDL events not applying to tables (CREATE DATABASE etc.).

      • getCurrentGtid

        String getCurrentGtid()

      • isLastSnapshot

        boolean isLastSnapshot()

      • getCurrentBinlogFilename

        String getCurrentBinlogFilename()

      • getCurrentBinlogPosition

        long getCurrentBinlogPosition()

      • getBinlogTimestampSeconds

        long getBinlogTimestampSeconds()

      • getCurrentRowNumber

        int getCurrentRowNumber()

      • createDocumentFromOffset

        public static Document createDocumentFromOffset​(Map<String,​?> offset)
        Create a Document from the given offset.
        Parameters:
        offset - the offset to create the document from.
        Returns:
        a Document with the offset data.

      • isPositionAtOrBefore

        public static boolean isPositionAtOrBefore​(Document recorded,
                                           Document desired,
                                           Predicate<String> gtidFilter)
        Determine whether the first offset is at or before the point in time of the second offset, where the offsets are given in JSON representation of the maps returned by offset().

        This logic makes a significant assumption: once a MySQL server/cluster has GTIDs enabled, they will never be disabled. This is the only way to compare a position with a GTID to a position without a GTID, and we conclude that any position with a GTID is *after* the position without.

        When both positions have GTIDs, then we compare the positions by using only the GTIDs. Of course, if the GTIDs are the same, then we also look at whether they have snapshots enabled.

        Parameters:
        recorded - the position obtained from recorded history; never null
        desired - the desired position that we want to obtain, which should be after some recorded positions, at some recorded positions, and before other recorded positions; never null
        gtidFilter - the predicate function that will return true if a GTID source is to be included, or false if a GTID source is to be excluded; may be null if no filtering is to be done
        Returns:
        true if the recorded position is at or before the desired position; or false otherwise