ndb - bug#27466 nf during nr can leave cluster in inconsistent state (recommit in 5.1)
Fix race condition between NODE_FAILREP and local INCL_NODEREQ loop Also retry on ZNODE_START_DISALLOWED_ERROR storage/ndb/include/kernel/signaldata/StartPerm.hpp: Move error code storage/ndb/src/kernel/blocks/ERROR_codes.txt: new error code storage/ndb/src/kernel/blocks/dbdih/Dbdih.hpp: Move error code storage/ndb/src/kernel/blocks/dbdih/DbdihMain.cpp: 1) retry also on ZNODE_START_DISALLOWED_ERROR 2) Change if() else in INCL_NODECONF to for-loop instead 3) (last but not least) fix bug, that could cause different block withing same node to have different opinion about node status solution is to check if node is still alive before sening next local INCL_NODEREQ storage/ndb/src/kernel/blocks/dbtc/DbtcMain.cpp: Add error insert to allow node to die during INCL_NODEREQ storage/ndb/src/kernel/blocks/suma/Suma.cpp: 1) let suma be well behaved (i.e reply to INCL_NODEREQ) 2) Add dump to print c_connceted_nodes/c_subscriber_nodes (8010) storage/ndb/test/ndbapi/testNodeRestart.cpp: new testcase storage/ndb/test/run-test/daily-basic-tests.txt: new testcase
Showing
Please register or sign in to comment