lot of data (10-20 terabyte), we want to move to linux. how should i do that?
Which type of replication is: master->slave or master -> slave1 (slave from master and master for slave2) -> slave2 (slave from slave2), or galera? MySQL, Percona, or MariaDB? Version MySQL?
It's a normal master-slave Normal mysql 5.5
This is version the end of life in December 2018; version 5.7 end of live in November 2023.... Maybe you need to upgrade this version? I don't think that you could correctly move to Linux with this version. You can try adding a node replica on Linux. Maybe you get problems with CHARACTER SET and COLLATION because these parameters maybe different on Windows and Linux. You need to check these parameters after adding node replica on Linux. Exemplary plan: 0. For example, your Master node name - WinMaster, slave1 node name - WinSlave1, slave2 node name - WinSlave2, slave3 node name - WinSlave3. 1. Install and setup MySQL 5.5 on a Linux node, for example, the node named LinSlave1. 2. You need to stop replication on WinSlave3 'STOP SLAVE SQL_THREAD;'. Important notice: you need to have a lot of disk space on this node for relay-log. 3. Take only the structure database into the file and maybe correct the parameters 'CHARACTER SET' and 'COLLATION' for the Linux in this file. 4. Export each table in a separate file; maybe for some tables you export more than one file. 5. Execute the script with the structure database on LinSlave1. 6. Download data from the data files to the database. 7. 'START SLAVE SQL_THREAD;' 8. Check the replication status on WinSlave3 to see if WinSlave3 caught up with the master - WinMaster. 9. Check the replication status on LinSlave1 to see if LinSlave1 caught up with the master - WinSlave3. If all is well, you make another node on Linux (slave from LinSlave1) and for some time you do LinSlave1 - master. You need to understand that this plan doesn't guarantee 100% success.
Why do you STOP SLAVE for WinSlave3? And to which node do you connect the LinSlave1 - to WinSlave3 or WinMaster? Also, why not export all tables in one dump? to avoid problems?
You have very big data tables. This way, you can reduce the load for your usual work. You'll be migrating from Windows to Linux.
I understand that data is huge but what exactly in this process will reduce the load? can you explain my other questions as well please? :)
step 2 - helped you reduce consequences if anything will break. The maximum problem for you is that you lost WinSlave3.
Обсуждают сегодня