Tags: gaizai/orchestrator
Tags
Supporting SQL_Delay - db column added - instance does not have problem if lag is within 10 seconds of sql_delay - visualizing sql_delay Reace conditions: - added mutex on cluster alias map MaxScale support: - move-up, move-below supports moving a slave up from maxscale, below a maxscale via simple repointing
This commit mainly targets two issues: 1. `SHOW BINARY LOGS` generates locks on mysql server or otherwise takes a long time to execute on busy servers with many binary logs 2. MultiMatchBelow/GetSlaveCandidate did not refresh slaves positions after StopSlavesNicely, hence elected wrong candidate breakdown: BinlogCoordinates: - added FileSmallerThan() (compares file-name only) - added NextFileCoordinates() -- works similarly to PreviousFileCoordinates() and comes to remove need of `SHOW BINARY LOGS` Instance: - GetNextBinaryLog() checks against SelfBinlogCoordinates for out-of-bounds binary log file (no need for `SHOW BINARY LOGS`) instance_dao.go: - not executing `SHOW BINARY LOGS` - StopSlavesNicely returns slice of refreshed slaves (positions updated after `STOP SLAVE` issued) - more debug messages all over the place instance_binlog_dao.go: - getLastPseudoGTIDEntryInInstance() uses heuristic iteration over binary logs - SearchPseudoGTIDEntryInInstance() uses heuristic iteration over binary logs - getNextBinlogEventsChunk() uses heuristic iteration over binary logs (correctly silently exiting upon end of binlogs) instance_topology.go: - removed redundant StopSlavesNicely on RegroupSlaves - sortedSlaves updates list of slaves with refreshed image - MultiMatchBelow updates list of slaves with refreshed image - more debug messages all over the place
PreviousNext