(mysql.info.gz) MySQL Cluster Limitations in 4
Info Catalog
(mysql.info.gz) MySQL Cluster Interconnects
(mysql.info.gz) NDBCluster
16.7 MySQL Cluster Limitations in 4.1
=====================================
Below is a list of known limitations with release 4.1 when comparing to
the storage engines MyISAM and InnoDB. Currently there are no plans to
address these in coming releases of 4.1 (but well in 5.0 or later
releases). At `http://bugs.mysql.com', category cluster, you fill find
known bugs which are intended to be fixed in upcoming releases of 4.1
(if marked 4.1). This list is intended to be complete with respect to
the above, please report discrepancies at `http://bugs.mysql.com',
category cluster. If this discrepancy will not be fixed in 4.1 it will
be added to this list.
Non compliance in syntax (resulting in error when running an existing application)
Not all charsets and collations supported, see MySQL
Cluster news-4.1.6 for a list of the ones supported.
No prefix indexes (can only index complete fields)
No text indexes
No geometry type support
Non compliance in limits/behavior (may result in error when running an existing application)
No partial rollback of transactions on error. As a result,
e.g. duplicate key error, will result in a rollback of
the whole transaction.
A number of hard limits exist which are configurable, but
available main memory in the cluster sets limits. See
complete list of configuration parameters MySQL
Cluster Config File. Most configuration parameters can be
upgraded on-line.
Database memory size and index memory size, `DataMemory'
and `IndexMemory' respectively.
How large transactions can be performed is set with
config parameter
`MaxNoOfConcurrentOperations' (bulk load,
truncate table and alter table are handled specially by
running several transactions, and thus does
not have this limitation)
Different limits related to tables and indexes, e.g. max
number of ordered indexes,
`MaxNoOfOrderedIndexes' etc.
Database names, table names and attribute names cannot be as
long in ndb tables as in other table handlers.
Attribute names will be truncated to 31 characters, and
if they are not unique after truncation gives an error.
Database name and table name can be max a total of 122
characters together.
Max number of meta data objects is limited to 1600 (includes
tables, system tables, indexes and BLOBs)
Max number of attributes per table limited to 128
Max row size of 8k (not including BLOBs)
Max number of attributes in a key is 32
Not supported features (no error, but not supported/enforced)
Foreign key construct is ignored (same behavior as MyISAM)
Savepoint and rollback to savepoint ignored (same behavior as
MyISAM)
Performance and limitations related
Query cache disabled since it is not invalidated if update
occurs on other MySQL server
Query performance issues due to sequential access to storage
engine and relatively more expensive to do many range
scans.
Records in range statistic not supported, giving non-optimal
query plans in some cases. Use constructs `USE INDEX' or
`FORCE INDEX' to work around non optimal query plans.
Unique hash index (constructed with `USING HASH') cannot be
used for accessing table if NULL is given as part of
the key.
Missing features
Only support for READ_COMMITTED isolation level (InnoDB
supports READ_COMMITTED, REPEATABLE_READ and SERIALIZABLE)
No disk durable commit (are replicated, but no guarantee that
logs are flushed to disk at commit)
Problems related to multiple MySQL servers (not related to MyISAM or InnoDB)
Alter table not fully blocking when running multiple MySQL
servers (no distributed table lock)
MySQL Replication will not work right off if updates are done
on several MySQL servers. If a partitioning scheme of
the database is done at the application level, and no
transactions takes place across these partitions, it
will work.
Cluster only related (not related to MyISAM or InnoDB)
All machines in the cluster must use the same architecture;
that is, all machines hosting nodes must be either
big-endian or little-endian, and you cannot use a
mixture of both. For example, you cannot have a
management node running on a PPC which directs a server
node that is running on an x86 machine. This restriction
does not apply to machines simply running `mysql' clients
that may be accessing the cluster.
No on-line schema changes (no on-line alter table, Ndb
on-line create index not accessible)
No on-line add/drop node
When using multiple management servers one must give nodes an
id explicitly in the connect strings since the
automatic allocation of node id's does not work across
several management servers.
When using multiple management servers one must take extreme
care to have the same configuration in all management
servers. No checks for this are performed.
Max number of storage nodes 48
Total max number of nodes 63 (MySQL Servers, storage nodes
and management servers)
Info Catalog
(mysql.info.gz) MySQL Cluster Interconnects
(mysql.info.gz) NDBCluster
automatically generated byinfo2html