What exactly is 'inconsistent' in the context of --no-locks and how does it interact with --use-savepoints?
It's pretty obvious that if mydumper --no-locks is used, MyISAM tables will not be consistent. But what if, "hypothetically speaking", I have a large (approaching 1TB) customer database in XtraDB Cluster 5.6 for which the customer data is entirely in InnoDB tables, with no MyISAM tables existing outside the mysql system schema, and the mysql schema can be considered static? Will --no-locks still give me all of the InnoDB data as a consistent single transaction? In this scenario, how does --no-locks interact with --use-savepoints?
In general, in this scenario in which there can be assumed to be no MyISAM data changes, all data is in InnoDB, and locks should be avoided at all costs if possible, what combination of options will give me a consistent snapshot of all of the InnoDB data with the least possible locking?
The invocation I'm testing out right now is as follows:
mydumper -Cce --use-savepoints --success-on-1146 --no-locks -L /root/mydumper.log -h [host] -o /[dump directory]
The --success-on-1146 is in use because this customer has thousands of small schemas with a high turnover rate; our current mysqldump-based dump script FREQUENTLY reports dump anomalies — sometimes several in a single dump cycle — because by the time mysqldump gets around to dumping a particular schema, it doesn't exist any more.
Question information
- Language:
- English Edit question
- Status:
- Solved
- Assignee:
- No assignee Edit question
- Solved by:
- Phil Stracchino
- Solved:
- Last query:
- Last reply: