Description:
Setup:
8 data nodes (eight separate machines)
2 mysql servers (two separate machines)
As this is table create phase there is no load on the system.
mysql server 51:
runs a script creating tables.
The first thing the script does is to drop a number of database.
mysql server 52:
connected to the cluster
When the script from 51 is executed and the databases are DROPPED we get:
For some databases we get this (completish log below):
081113 15:37:16 [ERROR] NDB Binlog: Skipping drop database 'bcd' since it contained local tables binlog schema event 'drop database bcd' from node 31.
081113 15:37:47 [ERROR] mysqld: Incorrect information in file: './tzf5001/cache_age.frm'
081113 15:37:47 [ERROR] mysqld: Incorrect information in file: './tzf5001/cache_age.frm'
081113 15:38:09 [ERROR] mysqld: Incorrect information in file: './tzf5001
081113 15:38:18 [ERROR] NDB Binlog: Skipping drop database 'tzf5001' since it contained local tables binlog schema event 'drop database tzf5001' from node 31.
What is "local tables binlog schema event"??
On server 51 we then get:
081113 15:39:59 [Note] NDB Binlog: CREATE TABLE Event: REPL$tms/tm_session_attribute
081113 15:40:01 [Note] NDB Binlog: CREATE TABLE Event: REPL$sdp5001_deploy/audit_info
081113 15:40:02 [Note] NDB Binlog: CREATE TABLE Event: REPL$sdp5001_deploy/deployment
081113 15:40:03 [Note] NDB create table: waiting max 119 sec for distributing ./sdp5001_deploy/deployment. epochs: (0/0,0/0,249519/9) injector proc_info: Waiting for event from ndbcluster
....
081113 15:43:54 [Note] NDB Binlog: CREATE TABLE Event: REPL$sdp5001_deploy/msisdn_job_log
081113 15:43:55 [Note] NDB create table: waiting max 119 sec for distributing ./sdp5001_deploy/msisdn_job_log. epochs: (0/0,0/0,249739/4) injector proc_info: Waiting for event from ndbcluster
081113 15:43:56 [Note] NDB create table: waiting max 118 sec for distributing ./sdp5001_deploy/msisdn_job_log. epochs: (0/0,0/0,249740/3) injector proc_info: Waiting for event from ndbcluster
081113 15:43:57 [Note] NDB create table: waiting max 117 sec for distributing ./sdp5001_deploy/msisdn_job_log. epochs: (0/0,0/0,249741/3) injector proc_info: Waiting for event from ndbcluster
081113 15:43:58 [Note] NDB create table: waiting max 116 sec for distributing ./sdp5001_deploy/msisdn_job_log. epochs: (0/0,0/0,249742/2) injector proc_info: Waiting for event from ndbcluster
081113 15:43:59 [Note] NDB create table: waiting max 115 sec for distributing ./sdp5001_deploy/msisdn_job_log. epochs: (0/0,0/0,249743/1) injector proc_info: Waiting for event from ndbcluster
081113 15:44:00 [Note] NDB create table: waiting max 114 sec for distributing ./sdp5001_deploy/msisdn_job_log. epochs: (0/0,0/0,249744/1) injector proc_info: Waiting for event from ndbcluster
081113 15:44:01 [Note] NDB create table: waiting max 113 sec for distributing ./sdp5001_deploy/msisdn_job_log. epochs: (0/0,0/0,249745/0) injector proc_info: Waiting for event from ndbcluster
After a while I abort the connection (kill server)
081113 15:44:02 [Warning] Aborted connection 3525 to db: 'sdp5001_deploy' user: 'root' host: 'localhost' (Got an error writing communication packets)
081113 15:44:03 [Warning] Aborted connection 3534 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error writing communication packets)
081113 15:44:04 [Warning] Aborted connection 3536 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error writing communication packets)
081113 15:44:05 [Warning] Aborted connection 3538 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error writing communication packets)
Now the tables are being created at normal speed again!
How to repeat:
will get scripts
Suggested fix:
will get scripts
