MySQL disk replication technology-DRBD: Comparison of advantages and disadvantages, precautions and best practices

MySQL disk replication technology-DRBD: advantages and disadvantages comparison, precautions and Best Practices DRBD is the block-level synchronous replication technology implemented by the kernel module. The synchronization level here can be adjusted because DRBD uses the NIC for block replication. If, using Infiniband for transmission can effectively process high concurrency. This is a type of replication storage, which is nothing more like a hot standby machine. Instead of storing HA, it is better to ensure data security in the industrial environment. it is more used on NFS servers and combined with Linux-HA projects, such as Packmaker and Heartbeat. Many people talk about DRBD split-brain and color-changing. Once used, we know that split-brain is not so easy to happen. do not set automatic start, manual start during restart, split-brain. In rare cases, DRBD is basically not restarted, and the server is not restarted. Basically, no split-brain problems are encountered. (I) advantages and disadvantages advantages ① no single point of failure (spof) risks exposed ② No cluster software faults, such as split-brain defects (1) the standby machine that wastes resources DRBD is not currently available for reading, so it is a waste of resources and everyone knows, production environment it is a pity that all the MySQL servers are configured with top-end servers. Therefore, it is a waste of machines online. (2) DRBD cannot replace the backup of DRBD, which may cause bad data. No matter what HA, data backup must be performed. yes, there is no backup, the whole HA is paralyzed, you are also fart, flash people (3) failover cannot be within seconds (4) is not very useful for MyISAM tables in fact, any system that expects high availability myISAM is not an option, because it takes a lot of time to fix the problem after MyISAM crash recovery, and data loss occurs. The write burden is increased mainly because InnoDB is set to innodb_flush_log_at_trx_commit = 1, resulting in many small and fsync () call, then, DRBD synchronization will be relatively slow (ii) note ① DRBD and Linux-HA version selection ② DRBD dedicated to using a network card ③ heartbeat network and vip service network separation ④ iptables for heartbeat Network open the ports of DRBD and Linux-HA, and enable the corresponding ports of VIP based on your application. 5. Monitor the status of DRBD and Linux-HA, make scripts, perform automatic detection, and issue alarms. 6. There are usually traffic Monitoring System (mrtg, cacti, etc) regular server load and traffic check (iii) Best practices DRBD only selects to copy the device that stores the binlog. If the active node fails, you can enable a log server on the passive node and then disable apply the binlog to all devices in the master database. Then, you can select one of the slave databases as the master database to replace the invalid system.