BF aborter hangs
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
MySQL patches by Codership |
Fix Released
|
Critical
|
Seppo Jaakola | ||
Trunk |
Fix Released
|
Critical
|
Seppo Jaakola |
Bug Description
Running sqlgen with default parameters against 2-node cluster makes one node hang in BF abort phase:
processlist form the failed node:
mysql> show processlist;
+-----+
| Id | User | Host | db | Command | Time | State | Info |
+-----+
| 1 | system user | | NULL | Sleep | 193 | invalidating query cache entries (table) | NULL |
| 2 | system user | | NULL | Sleep | 60240 | wsrep aborter idle | NULL |
| 726 | test | devel.codership
| 729 | test | devel.codership
| 730 | test | devel.codership
| 731 | root | localhost:52980 | NULL | Query | 0 | NULL | show processlist |
+-----+
6 rows in set (0.00 sec)
Error log from the failed node:
WSREP: BF victim is waitingWSREP c_lock is waiting: 21928WSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait longWSREP: BF lock wait long
Here is how to reproduce the issue:
1. Start two nodes (abyssinian and bengal) in this case
2. create sqlgen database with: (credentials omitted):
sqlgen --host=abyssinian --create=1 --tables=4 --rows=4 --duration=0 --users=1
3. start sqlgen load with 4 concurrent users:
sqlgen --host=abyssinian --host=bengal --create=0 --duration=60
4. monitor processlists in both nodes