|
Consider the situation. CS cluster has N PM modules with M dbroots each. On PM went down so we have N-1*M dbroots with data. With current version of CS this data is currently inaccessable. This could be reproduced using official CS Zeppelin images. There are at least two errors reported when one of PM-s in 2UM2PM scenario went down:
- ERROR 1815 (HY000): Internal error: InetStreamSocket::readToMagic: Remote is closed
- ERROR 1815 (HY000): Internal error: st: 0 TupleBPS::sendPrimitiveMessages() caught an exception: IDB-2034: At least one DBRoot required for that query is offline.
One should investigate the reason behind the errors and enable partial data availability.
|