MySQL - Replication FAQ

Why do I sometimes see more than one Binlog_Dump thread on the master after I have restarted the slave?

Binlog_Dump is a continuous process that is handled by the server in the following way:

Catch up on the updates.

Once there are no more updates left, go into pthread_cond_wait(), from which we can be awakened either by an update or a kill.

On wake up, check the reason. If we are not supposed to die, continue the Binlog_dump loop.

If there is some fatal error, such as detecting a dead client, terminate the loop.

So if the slave thread stops on the slave, the corresponding Binlog_Dump thread on the master will not notice it until after at least one update to the master (or a kill), which is needed to wake it up from pthread_cond_wait(). In the meantime, the slave could have opened another connection, which resulted in another Binlog_Dump thread.

The above problem should not be present in Version 3.23.26 and later versions. In Version 3.23.26 we added server-id to each replication server, and now all the old zombie threads are killed on the master when a new replication thread connects from the same slave

How do I rotate replication logs?

In Version 3.23.28 you should use PURGE MASTER LOGS TO command after determining which logs can be deleted, and optionally backing them up first. In earlier versions the process is much more painful, and cannot be safely done without stopping all the slaves in the case that you plan to re-use log names. You will need to stop the slave threads, edit the binary log index file, delete all the old logs, restart the master, start slave threads,and then remove the old log files.

How do I upgrade on a hot replication setup?

If you are upgrading pre-3.23.26 versions, you should just lock the master tables, let the slave catch up, then run FLUSH MASTER on the master, and FLUSH SLAVE on the slave to reset the logs, then restart new versions of the master and the slave. Note that the slave can stay down for some time - since the master is logging all the updates, the slave will be able to catch up once it is up and can connect.

After 3.23.26, we have locked the replication protocol for modifications, so you can upgrade masters and slave on the fly to a newer 3.23 version and you can have different versions of MySQL running on the slave and the master, as long as they are both newer than 3.23.26.

What issues should I be aware of when setting up two-way replication?

MySQL replication currently does not support any locking protocol between master and slave to guarantee the atomicity of a distributed (cross-server) update. In 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 will make the update of client A work differently than it did on co-master 1. Thus when the update of client A will make it to co-master 2, it will produce tables that will be different than what you have on co-master 1, even after all the updates from co-master 2 have also propagated. So you should not co-chain two servers in a two-way replication relationship, unless you are sure that you updates can safely happen in any order, or unless you take care of mis-ordered updates somehow in the client code.

You must also realize that two-way replication actually does not improve performance very much, if at all, as far as updates are concerned. Both servers need to do the same amount of updates each, as you would have one server do. The only difference is that there will be a little less lock contention, because the updates originating on another server will be serialized in one slave thread. This benefit, though, might be offset by network delays.

How can I use replication to improve performance of my system?

You should set up one server as the master, and direct all writes to it, and configure as many slaves as you have the money and rackspace for, distributing the reads among the master and the slaves. You can also start the slaves with --skip-bdb, --low-priority-updates and --delay-key-write-for-all-tables to get speed improvements for the slave. In this case the slave will use non-transactional MyISAM tables instead of BDB tables to get more speed.

What should I do to prepare my client code to use performance-enhancing replication?

A: If the part of your code that is responsible for database access has been properly abstracted/modularized, converting it to run with the replicated setup should be very smooth and easy - just change the implementation of your database access to read from some slave or the master, and to awlays write to the master. If your code does not have this level of abstraction, setting up a replicated system will give you an opportunity/motivation to it clean up. You should start by creating a wrapper library /module with the following functions:

safe_writer_connect()

safe_reader_connect()

safe_reader_query()

safe_writer_query()

safe_ means that the function will take care of handling all the error conditions.

You should then convert your client code to use the wrapper library. It may be a painful and scary process at first, but it will pay off in the long run. All applications that follow the above pattern will be able to take advantage of one-master/many slaves solution. The code will be a lot easier to maintain, and adding troubleshooting options will be trivial. You will just need to modify one or two functions, for example, to log how long each query took, or which query, among your many thousands, gave you an error. If you have written a lot of code already, you may want to automate the conversion task by using Monty's replace utility, which comes with the standard distribution of MySQL, or just write your own Perl script. Hopefully, your code follows some recognizable pattern. If not, then you are probably better off re-writing it anyway, or at least going through and manually beating it into a pattern.

Note that, of course, you can use different names for the functions. What is important is having unified interface for connecting for reads, connecting for writes, doing a read, and doing a write.

When and how much can MySQL replication improve the performance of my system?

MySQL replication is most beneficial for a system with frequent reads and not so frequent writes. In theory, by using a one master/many slaves setup you can scale 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.

In order to determine how many slaves you can get before the added benefits begin to level out, and how much you can improve performance of your site, you need to know your query patterns, and empirically (by benchmarking) determine the relationship between the throughput on reads (reads per second, or max_reads) and on writes max_writes) on a typical master and a typical slave. The example below will show you a rather simplified calculation of what you can get with replication for our imagined system.

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

reads = 1200 - 2 * writes (from bencmarks)

reads = 9* writes / (N + 1) (reads split, but writes go to all servers)

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

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

So if N = 0, which means we have no replication, our system can handle 1200/11, about 109 writes per second (which means we will have 9 times as many reads due to the nature of our application).

If N = 1, we can get up to 184 writes per second.

If N = 8, we get up to 400.

If N = 17, 480 writes.

Eventually as N approaches infinity (and our budget negative infinity), we can get very close to 600 writes per second, increasing system throughput about 5.5 times. However, with only 8 servers, we increased it almost 4 times already.

Note that our computations assumed infinite network bandwidth, and neglected several other factors that could turn out to be signficant on your system. In many cases, you may not be able to make a computation similar to the one above that will accurately predict what will happen on your system if you add N replication slaves. However, answering the following questions should help you decided whether and how much, if at all, the replication will improve the performance of your system:

What is the read/write ratio on your system?

How much more write load can one server handle if you reduce the reads?

How many slaves do you have bandwidth for on your network?

How can I use replication to provide redundancy/high availability?

With the currently available features, you would have to set up a master and a slave (or several slaves), and write a script that will monitor the master to see if it is up, and instruct your applications and the slaves of the master change in case of failure. Some suggestions:

To tell a slave to change the master use the CHANGE MASTER TO command.

A good way to keep your applications informed where the master is by having a dynamic DNS entry for the master. With bind you can use nsupdate to dynamically update your DNS.

You should run your slaves with the log-bin option and without log-slave-updates. This way the slave will be ready to become a master as soon as you issue STOP SLAVE; RESET MASTER, and CHANGE MASTER TO on the other slaves. It will also help you catch spurious updates that may happen because of misconfiguration of the slave (ideally, you want to configure access rights so that no client can update the slave, except for the slave thread) combined with the bugs in your client programs (they should never update the slave directly).

We are currently working on intergrating an automatic master election system into MySQL, but until it is ready, you will have to create your own monitoring tools.