A.13 MySQL 8.0 FAQ: Replication

No, it does not. The slave can go down or stay disconnected for hours or even days, and then reconnect and catch up on updates. For example, you can set up a master/slave relationship over a dial-up link where the link is up only sporadically and for short periods of time. The implication of this is that, at any given time, the slave is not guaranteed to be in synchrony with the master unless you take some special measures.

To ensure that catchup can occur for a slave that has been disconnected, you must not remove binary log files from the master that contain information that has not yet been replicated to the slaves. Asynchronous replication can work only if the slave is able to continue reading the binary log from the point where it last read events.

Yes, networking must be enabled on the master and slave. If networking is not enabled, the slave cannot connect to the master and transfer the binary log. Check that the skip-networking option has not been enabled in the configuration file for either server.

Check the Seconds_Behind_Master column in the output from SHOW SLAVE STATUS. See Section 17.1.7.1, “Checking Replication Status”.

When the slave SQL thread executes an event read from the master, it modifies its own time to the event timestamp. (This is why TIMESTAMP is well replicated.) In the Time column in the output of SHOW PROCESSLIST, the number of seconds displayed for the slave SQL thread is the number of seconds between the timestamp of the last replicated event and the real time of the slave machine. You can use this to determine the date of the last replicated event. Note that if your slave has been disconnected from the master for one hour, and then reconnects, you may immediately see large Time values such as 3600 for the slave SQL thread in SHOW PROCESSLIST. This is because the slave is executing statements that are one hour old. See Section 17.2.2, “Replication Implementation Details”.

Use the following procedure:

MySQL replication currently does not support any locking protocol between master and slave to guarantee the atomicity of a distributed (cross-server) update. In other words, it is possible for client A to make an update to co-master 1, and in the meantime, before it propagates to co-master 2, client B could make an update to co-master 2 that makes the update of client A work differently than it did on co-master 1. Thus, when the update of client A makes it to co-master 2, it produces tables that are different from what you have on co-master 1, even after all the updates from co-master 2 have also propagated. This means that you should not chain two servers together in a two-way replication relationship unless you are sure that your updates can safely happen in any order, or unless you take care of mis-ordered updates somehow in the client code.

You should also realize that two-way replication actually does not improve performance very much (if at all) as far as updates are concerned. Each server must do the same number of updates, just as you would have a single server do. The only difference is that there is a little less lock contention because the updates originating on another server are serialized in one slave thread. Even this benefit might be offset by network delays.

Set up one server as the master and direct all writes to it. Then configure as many slaves as you have the budget and rackspace for, and distribute the reads among the master and the slaves. You can also start the slaves with the --skip-innodb and --low-priority-updates options, and set the delay_key_write system variable to ALL to get speed improvements on the slave end. In this case, the slave uses nontransactional MyISAM tables instead of InnoDB tables to get more speed by eliminating transactional overhead.

See the guide to using replication as a scale-out solution, Section 17.3.5, “Using Replication for Scale-Out”.

MySQL replication is most beneficial for a system that processes frequent reads and infrequent writes. In theory, by using a single-master/multiple-slave setup, you can scale the system by adding more slaves until you either run out of network bandwidth, or your update load grows to the point that the master cannot handle it.

To determine how many slaves you can use before the added benefits begin to level out, and how much you can improve performance of your site, you must know your query patterns, and determine empirically by benchmarking the relationship between the throughput for reads and writes on a typical master and a typical slave. The example here shows a rather simplified calculation of what you can get with replication for a hypothetical system. Let reads and writes denote the number of reads and writes per second, respectively.

Let's say that system load consists of 10% writes and 90% reads, and we have determined by benchmarking that reads is 1200 - 2 * writes. In other words, the system can do 1,200 reads per second with no writes, the average write is twice as slow as the average read, and the relationship is linear. Suppose that the master and each slave have the same capacity, and that we have one master and N slaves. Then we have for each server (master or slave):

reads = 1200 - 2 * writes

reads = 9 * writes / (N + 1) (reads are split, but writes replicated to all slaves)

9 * writes / (N + 1) + 2 * writes = 1200

writes = 1200 / (2 + 9/(N + 1))

The last equation indicates the maximum number of writes for N slaves, given a maximum possible read rate of 1,200 per second and a ratio of nine reads per write.

This analysis yields the following conclusions:

These computations assume infinite network bandwidth and neglect several other factors that could be significant on your system. In many cases, you may not be able to perform a computation similar to the one just shown that accurately predicts what will happen on your system if you add N replication slaves. However, answering the following questions should help you decide whether and by how much replication will improve the performance of your system:

How you implement redundancy is entirely dependent on your application and circumstances. High-availability solutions (with automatic failover) require active monitoring and either custom scripts or third party tools to provide the failover support from the original MySQL server to the slave.

To handle the process manually, you should be able to switch from a failed master to a pre-configured slave by altering your application to talk to the new server or by adjusting the DNS for the MySQL server from the failed server to the new server.

For more information and some example solutions, see Section 17.3.8, “Switching Masters During Failover”.

Check the value of the binlog_format system variable:

The value shown will be one of STATEMENT, ROW, or MIXED. For MIXED mode, statement-based logging is used by default but replication switches automatically to row-based logging under certain conditions, such as unsafe statements. For information about when this may occur, see Section 5.4.4.3, “Mixed Binary Logging Format”.

Slaves automatically know which format to use.

Start the server with the --replicate-wild-ignore-table=mysql.% option to ignore replication for tables in the mysql database.

Yes.

Yes.