MySQL Replibeertion

MySQL Replibeertion was the last scheduled session on Day 2, but not withstanding there was free beer (a lot of), there was a serious side with a Replication Discussion.

One of the first questions by Jeremy was “Are the any big replication users?” to which Sheeri quickly replied “Are you calling me fat again”.

This was a highly interactive session, here are some of the points from the audience.

Some Uses of Replication

  • Backup
  • Hot standby
  • Scaling
  • Data Warehousing
    • Slaves are larger then your database
  • For no special reason
  • Consolidation of multiple sources
  • Support for multiple indexes

Issues

  • Can break
  • replication lag
  • bi-directional replication is not supported
  • hard to setup replication/initialization upto point to run one command
  • hard to know when the slave is out of sync (working but broken) diagnose
  • does not manage binary logs for you (max-bin-logs )
  • no row level replication (5.1 row based replication, change based replication, good and bad )
  • Serialized execution on the slave
  • Master does not keep track of the slaves (to the master, the slave is just another connection) Jeremy comment “it really really sucks in production systems.”
  • No multi-master replication. A slave can not have more then one master.
  • ring replication. No idea when something breaks what is right.
  • no ability for delayed duplication.
  • no way to get binlogs back. (manually twink the info file)
  • Master doesn’t care what data is on the slave.
  • A replication backup is really only good for the restoration of that machine
  • Default reconnection timeout is way, way to high (default of 1 minute). It should be at most 1 second with exponential fallback. (master-connect-retry), no fallback, no max number of retries, logged in error log every time.

Feature Results (Things replication needs, what you want to see)

  • Delayed Replication
  • Registered Salves in the Master
  • Import Binlog
  • Checksum Table Events (Need ability for table checksum to be added to binlog periodically so it can be checked by the slave.)
  • Global Sequence Number
  • Connect Retry Exponential Backoff
  • Hetrogous Replication (Oracle to MySQL). Golden Gate Software has a commercial offering
  • Command Exclusion List (sql_log_bin=0)
  • Replication filters by data on the slave
  • Show upcoming queries, skip query
  • Multi-Master to one slave
  • piping mysqlbin log commands into the mysql client fails for some character sets.
  • Binlog index capability
  • Checksum of Binary Events to determine a command is valid
  • command line interface in mysqllog so you could go backwards and forwards, then execute commands.

Check out more at Google Code Blog.

Tagged with: Databases General MySQL MySQL Camp 1 - 2006

Related Posts

More CPUs or Newer CPUs

In a CPU-bound database workload, regardless of price, would you scale-up or scale-new? What if price was the driving factor, would you scale-up or scale-new? I am using as a baseline the first available AWS Graviton2 processor for RDS (r6g).

Read more

An Interesting Artifact with AWS RDS Aurora Storage

As part of using public datasets with my own Benchmarking Suite I wanted upsize a dataset for larger volume testing. I have always used the INFORMATION_SCHEMA.TABLES data_length and index_length columns as a sufficiently accurate measurement for actual disk space used.

Read more

How long does it take the ReadySet cache to warm up?

During my setup of benchmarking I run a quick test-sysbench script to ensure my configuration is right before running an hour+ duration test. When pointing to a Readyset cache where I have cached the 5 queries used in the sysbench test, but I have not run any execution of the SQL, throughput went up 10x in 5 seconds.

Read more