I have a logging table on the master server that is inserted into very often. I don't need this table replicated to the slave servers, and in fact I already have replicate-ignore-table set on the slaves to ignore it.
However, that only happens after all of those inserts are fetched from the master. I'd like to prevent those inserts from getting sent to the slaves entirely for 2 reasons:
Cut down on network traffic between the servers
I've had cases of the relay log entries being corrupted (and having to skip corrupted entries). Given the quantity of inserts into the logging table, it's always on those inserts (which aren't necessary anyway).
Is it possible to somehow prevent the master from sending back the logs for a specific table? Or, prevent the inserts from showing up in the master's bin-log files? I'm only aware of ignoring databases in the master's bin-log files.
Thanks.
In your code, send "SET SESSION sql_log_bin=0" to MySQL before inserting a logging row. Then set it back to 1 afterward.
This approach gives you fine-grained control over when and when not to binary-log. Only possible drawback is that the database user will need the SUPER privilege.
Related
In MySQL master-slave synchronization, we configured the ignore table because of the performance problem of the slave. However, since the configuration is configured on the slave, when the host writes data in large quantities, will the data be synchronized to the slave and then filtered by the slave? In fact, if you ignore the table configuration, will the performance improvement be minimal? Or does it mean that when this data is written to the host, the ignore table information of the slave is read? Data transmission will not occur, which can greatly reduce the load of the slave? After all, I configure the ignore table to avoid the performance problems of the slave. If a large amount of data is written, the slave can't carry it.
Create a new database; I'll call it NoRepl
Put binlog-ignore-db=NoRepl on the Primary.
Move that big table into the database NoRepl.
That way, the filtering will occur much sooner -- thereby not clogging up replication either on the Replica or in the Primary's binlog.
If you are using Primary-Primary or you have another Replica that does want to see this table, then my suggestion will not work.
I have one master and two slaves.
Is it possible to restrict a particular read-only user to query only against the second slave (disallowing him from running any queries on the master and the first slave)?
I see that one can do the following to make un-replicated changes to the master, but what I think I need is to make changes to one slave and not the other.
SET sql_log_bin = {OFF|ON}
And the GRANT syntax allows one to limit what host users come from, but -- as far as I understand -- not which DB server(s) the person can use.
I didn't find much in a web search -- perhaps that's a hint that there's a better way to solve this problem. Basically I'm asking if this can be enforced by the database since the restriction I want applies to just this one user.
For context: a slave is basically just a server that copies every action that happened on the master. Depending on your configuration, the slaves will either just run the same queries that have been executed on the master, or apply a list of changes to individual rows to the slave.
To add a user just for a specific slave, you can do this directly on the slave. Anything you do here will only affect this slave. If your user currently exists on the master (and slaves), you would first have to drop him/remove his permissions, wait until this change has been replicated to all the slaves (which might also depend on your configuration), then add/modify this user directly on the slave.
You may need to temporarily disable a read_only or super_read_only setting (on the slave), which exists to prevent accidently executing something on the slave - but that is what you want to do.
Since your slave now deviates (slightly) from the master, if you would now run a query that alters that user on your master (e.g. drop it again), it might have a different effect on the master and the slave. This will depend on your configuration, but keep it in mind.
I'm currently thinking about the following problem:
A customer has set up a simple master/slave replication between two mariaDB systems. For unknown reasons they have set the flag "Replicate_Wild_Ignore_Table" to skip "logdb.%". Obviously, they decided to skip the skipping of that database and want the logdb to be included in the replication again.
I'm curious now, is it possible to somehow remove that flag and have the database in question be replicated as the rest or is there no way to circumvent the "stop slave, dump master, import dump, recreate replication based on current logpos, start slave" procedure?
You can't assume that the master still has all relevant binlogs that once contained updates to the logdb.% tables. That is, even if you could re-apply those updates, do you have enough history to account for all changes to the tables?
Another risk is if you use statement-based replication, if there were ever statements that referenced both a table in logdb.% and a table in another database, the replication filter has skipped that statement. So for example:
INSERT INTO mydb.mytable SELECT * FROM logdb.othertable;
Therefore even the tables that are not in logdb.% might be compromised. The point is you don't know for sure.
The bottom line is that you should definitely reinitialize the replica now by taking a current backup of the master, and avoid using replication filters in the future.
If you use InnoDB tables, you might consider using Percona XtraBackup to make the process easier. See https://www.percona.com/doc/percona-xtrabackup/2.3/howtos/setting_up_replication.html
With a replicating mysql master to master database with innodb engine, if one transaction were to initiate on database A will that row lock for database B until the transaction has been committed?
The master getting the first transaction is completely separate from the second master and they communicate through a binary log.
https://dev.mysql.com/doc/refman/5.7/en/replication-formats.html
In the case of something requiring a transaction, then the actual statements are not written to the log until the transaction is complete.
https://dev.mysql.com/doc/refman/5.7/en/replication-features-transactions.html
So the second master should be completely unhindered, since it won't actually know anything about the request until the first master is done processing it.
(Standard caveats though of it may depend on what type of replication SBR/RBR/mix and the actual transactions.)
mysql ignore-db works according to server my.cnf AFAIK,
i.e.
binlog-ignore-db = mysql
replicate-ignore-db = mysql
I am not sure, if this works from client side too, can anyone explain the mechanism, how can i be able to send from master but not accept in client side.
Why i want to do this? I have multiple slave "2 slave" must replicate MySQL table where as in other 2 should not be overwriten. Where as every other table will be replicated.
Reading this: http://dev.mysql.com/doc/refman/5.6/en/replication-rules-db-options.html didnt make me clear enough.
binlog-ignore-db is a master-side setting, it tells the Master not to log changes taking place on the listed DB.
replicate-ignore-db is a slave-side setting, it tells the Slave to ignore incoming log information related to the listed DB
The typical use case is when you want to replicate different databases from one single Master to different Slaves. The Master must log all changes occurring in all databases (minus those possibly excluded by binlog-ignore-db, i.e. database that will not be replicated anywhere).
Each Slave will receive the full binary log, but will only replicate changes related to the selected databases (i.e. databases not excluded by replicate-ignore-db -- this list would be different on each Slave).
(mysql database being a system database, it should be ignored from both ends, unless you really, really really know what you are doing).