17.7.2 Group Replication Limitations

The following known limitations exist for Group Replication. Note that the limitations and issues described for multi-primary mode groups can also apply in single-primary mode clusters during a failover event, while the newly elected primary flushes out its applier queue from the old primary.

Tip

Group Replication is built on GTID based replication, therefore you should also be aware of Section 16.1.3.6, “Restrictions on Replication with GTIDs”.

• Gap Locks.  The certification process does not take into account gap locks, as information about gap locks is not available outside of InnoDB. See Gap Locks for more information.

  Note

  Unless you rely on REPEATABLE READ semantics in your applications, we recommend using the READ COMMITTED isolation level with Group Replication. InnoDB does not use gap locks in READ COMMITTED, which aligns the local conflict detection within InnoDB with the distributed conflict detection performed by Group Replication.

• Table Locks and Named Locks.  The certification process does not take into account table locks (see Section 13.3.5, “LOCK TABLES and UNLOCK TABLES Syntax”) or named locks (see GET_LOCK()).

• Replication Event Checksums.  Due to a design limitation of replication event checksums, Group Replication cannot currently make use of them. Therefore set --binlog-checksum=NONE.

• SERIALIZABLE Isolation Level.  SERIALIZABLE isolation level is not supported in multi-primary groups by default. Setting a transaction isolation level to SERIALIZABLE configures Group Replication to refuse to commit the transaction.

• Concurrent DDL versus DML Operations.  Concurrent data definition statements and data manipulation statements executing against the same object but on different servers is not supported when using multi-primary mode. During execution of Data Definition Language (DDL) statements on an object, executing concurrent Data Manipulation Language (DML) on the same object but on a different server instance has the risk of conflicting DDL executing on different instances not being detected.

• Foreign Keys with Cascading Constraints.  Multi-primary mode groups (members all configured with group_replication_single_primary_mode=OFF) do not support tables with multi-level foreign key dependencies, specifically tables that have defined CASCADING foreign key constraints. This is because foreign key constraints that result in cascading operations executed by a multi-primary mode group can result in undetected conflicts and lead to inconsistent data across the members of the group. Therefore we recommend setting group_replication_enforce_update_everywhere_checks=ON on server instances used in multi-primary mode groups to avoid undetected conflicts.

  In single-primary mode this is not a problem as it does not allow concurrent writes to multiple members of the group and thus there is no risk of undetected conflicts.

• MySQL Enterprise Audit and MySQL Enterprise Firewall.  Prior to version 5.7.21 MySQL Enterprise Audit and MySQL Enterprise Firewall use MyISAM tables in the mysql system database. Group Replication does not support MyISAM tables.

• Multi-primary Mode Deadlock.  When a group is operating in multi-primary mode, SELECT .. FOR UPDATE statements can result in a deadlock. This is because the lock is not shared across the members of the group, therefore the expectation for such a statement might not be reached.

• Replication Filters.  Replication filters cannot be used on a MySQL server instance that is configured for Group Replication, because filtering transactions on some servers would make the group unable to reach agreement on a consistent state.

Limit on Group Size

The maximum number of MySQL servers that can be members of a single replication group is 9. If further members attempt to join the group, their request is refused. This limit has been identified from testing and benchmarking as a safe boundary where the group performs reliably on a stable local area network.

Limits on Transaction Size

If an individual transaction results in message contents which are large enough that the message cannot be copied between group members over the network within a 5-second window, members can be suspected of having failed, and then expelled, just because they are busy processing the transaction. Large transactions can also cause the system to slow due to problems with memory allocation. To avoid these issues use the following mitigations:

• Where possible, try and limit the size of your transactions. For example, split up files used with LOAD DATA into smaller chunks.

• Use the system variable group_replication_transaction_size_limit to specify a maximum transaction size that the group will accept. In MySQL 5.7, this system variable defaults to zero, but in MySQL 8.0, it defaults to a maximum transaction size of 150000000 bytes (approximately 143 MB). Transactions above this limit are rolled back and are not sent to Group Replication's Group Communication System (GCS) for distribution to the group. Adjust the value of this variable depending on the maximum message size that you need the group to tolerate, bearing in mind that the time taken to process a transaction is proportional to its size.

• Use the system variable group_replication_compression_threshold to specify a message size above which compression is applied. This system variable defaults to 1000000 bytes (1 MB), so large messages are automatically compressed. Compression is carried out by Group Replication's Group Communication System (GCS) when it receives a message that was permitted by the group_replication_transaction_size_limit setting but exceeds the group_replication_compression_threshold setting. If you set the system variable value to zero, compression is deactivated. For more information, see Section 17.9.7.2, “Message Compression”.

If you have deactivated message compression and do not specify a maximum transaction size, the upper size limit for a message that can be handled by the applier thread on a member of a replication group is the value of the member's slave_max_allowed_packet system variable, which has a default and maximum value of 1073741824 bytes (1 GB). A message that exceeds this limit fails when the receiving member attempts to handle it. The upper size limit for a message that a group member can originate and attempt to transmit to the group is 4294967295 bytes (approximately 4 GB). This is a hard limit on the packet size that is accepted by the group communication engine for Group Replication (XCom, a Paxos variant), which receives messages after GCS has handled them. A message that exceeds this limit fails when the originating member attempts to broadcast it.