hcq/MaxScale
hcq/MaxScale
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Documentation cleanup for Limitations.md

Browse Source 
Headers are now more consistent. Reordered headers. Changed wording.
This commit is contained in:
Esa Korhonen 2017-01-30 17:32:17 +02:00
parent ea0b4d6b45
commit 6db1cc25af
1 changed files with 139 additions and 195 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

334
Documentation/About/Limitations.md
View File

@ -1,182 +1,177 @@
# Limitations and Known Issues within MariaDB MaxScale
The purpose of this documentation is to provide a central location that
will document known issues and limitations within the MariaDB MaxScale product and
the plugins that form part of that product. Since limitations may related
to specific plugins or to MariaDB MaxScale as a whole this document is divided
into a number of sections, the purpose of which are to isolate the
limitations to the components which illustrate them.
This document lists known issues and limitations in MariaDB MaxScale and its
plugins. Since limitations are related to specific plugins, this document is
divided into several sections.
# Limitations in the MariaDB MaxScale core
## Protocol limitations
This section describes the limitations that are common to all
configuration of plugins with MariaDB MaxScale.
### Limitations with MySQL Protocol support (MySQLClient)
## Crash if one of several listeners for a Service fails as startup
Compression is not included in the MySQL server handshake.
If a service has multiple listeners and one of those listeners fails
at startup, MariaDB MaxScale will crash.
## Authenticator limitations
A typical reason for a listener to fail is that it has been configured
with a non-existing socket path or a port that MariaDB MaxScale is not allowed
to use.
### Limitations in the GSSAPI authenticator
Workaround: Ensure that socket paths and ports are valid.
Issues [MXS-710](https://jira.mariadb.org/browse/MXS-710) and
[MXS-711](https://jira.mariadb.org/browse/MXS-711) relate to this.
# Protocol limitations
## Limitations with MySQL Protocol support (MySQLClient)
Compression is not included in MySQL server handshake
# Authenticator limitations
## Limitations in the GSSAPI authentication
Currently MaxScale only supports GSSAPI authentication when the backend
Currently, MariaDB MaxScale only supports GSSAPI authentication when the backend
connections use GSSAPI authentication. Client side GSSAPI authentication with a
different backend authentication module is not supported.
# Monitor limitations
### Limitations in the MySQL authenticator (MySQLAuth)
* MariaDB MaxScale supports authentication that uses wildcard matching in
hostnames in the `mysql.user` table of the backend database. For IP address
entries either `%` or `_`-wildcards are accepted, they should not be mixed in
the same entry. For text addresses both wildcards can be mixed.
* Wildcards in text-form hostnames are not supported.
* MySQL old style passwords are not supported. MySQL versions 4.1 and newer use a
new authentication protocol which does not support pre-4.1 style passwords.
* When users have different passwords based on the host from which they connect
MariaDB MaxScale is unable to determine which password it should use to connect
to the backend database. This results in failed connections and unusable
usernames in MariaDB MaxScale.
## Filter limitations
Filters are not guaranteed to receive complete MySQL packets if they are used
with the readconnroute router. This can be fixed by using the readwritesplit
router.
### Database Firewall limitations (dbfwfilter)
The Database Firewall filter does not support multi-statements. Using them will
result in an error being sent to the client.
## Monitor limitations
A server can only be monitored by one monitor. If multiple monitors monitor the
same server, the state of the server is non-deterministic.
## Limitations with Galera Cluster Monitoring (galeramon)
### Limitations with Galera Cluster Monitoring (galeramon)
The default master selection is based only on MIN(wsrep_local_index). This
can be influenced with the server priority mechanic described in the
[Galera Monitor](../Monitors/Galera-Monitor.md) manual.
# Router limitations
## Router limitations
## Limitations in the connection router (readconnroute)
### Avrorouter limitations (avrorouter)
* If Master changes (ie. new Master promotion) during current connection
the router cannot check the change.
The avrorouter does not support the following data types and conversions:
* Sending of binary data with LOAD DATA LOCAL INFILE is not supported
* BIT
* Fields CAST from integer types to string types
## Limitations in the Read/Write Splitter (readwritesplit)
The avrorouter does not do any crash recovery. This means that the avro files
need to be truncated to valid block lengths before starting the avrorouter.
### Limitations in the connection router (readconnroute)
If Master changes (ie. new Master promotion) during current connection, the
router cannot check the change.
Sending of binary data with `LOAD DATA LOCAL INFILE` is not supported.
### Limitations in the Read/Write Splitter (readwritesplit)
Read queries are routed to the master server in the following situations:
* if they are executed inside an open transaction
* query is executed inside an open transaction
* query is a prepared statement
* statement includes a stored procedure or an UDF call
* if there are multiple statements inside one query e.g. `INSERT INTO ... ; SELECT
LAST_INSERT_ID();`
* in case of prepared statement execution
#### Backend write timeout handling
* statement includes a stored procedure, or an UDF call
* if there are multiple statements inside one query e.g.
`INSERT INTO ... ; SELECT LAST_INSERT_ID();`
### Backend write timeout handling
The backend connections opened by the readwritesplit will not be kept alive if
they aren't used. To keep all of the connections alive, a session command must
be periodically executed (for example `SET @a = 1;`).
The backend connections opened by readwritesplit will not be kept alive if they
aren't used. To keep all of the connections alive, a session command must be
periodically executed (for example `SET @a = 1;`).
If the backend server is configured with a low
[wait_timeout](https://mariadb.com/kb/en/mariadb/server-system-variables/#wait_timeout),
it is possible that connections get closed during long sessions. It is
recommended to set the wait_timeout to a high value and let MaxScale handle the
client timeouts. This can be achieved by using the
recommended to set the `wait_timeout` to a high value and let MariaDB MaxScale
handle the client timeouts. This can be achieved by using the
[_connection_timeout_](../Getting-Started/Configuration-Guide.md#connection_timeout)
parameter for the service.
### Limitations in multi-statement handling
#### Limitations in multi-statement handling
When a multi-statement query is executed through the readwritesplit
router, it will always be routed to the master. With the default
configuration, all queries after a multi-statement query will be routed to
the master to prevent possible reads of false data.
When a multi-statement query is executed through the readwritesplit router, it
will always be routed to the master. With the default configuration, all queries
after a multi-statement query will be routed to the master to prevent possible
reads of false data.
You can override this behavior with the `strict_multi_stmt=false` router
option. In this mode, the multi-statement queries will still be routed to
the master but individual statements are routed normally. If you use
multi-statements and you know they don't modify the session state in any
relevant way, you can disable this option for better performance.
You can override this behavior with the `strict_multi_stmt=false` router option.
In this mode, the multi-statement queries will still be routed to the master but
individual statements are routed normally. If you use multi-statements and you
know they don't modify the session state in any relevant way, you can disable
this option for better performance.
For more information, read the
[ReadWriteSplit](../Routers/ReadWriteSplit.md) router documentation.
### Parsing related limitations
#### Parsing limitations
Galera Cluster variables, such as @@wsrep_node_name, are not resolved by
the embedded MariaDB parser. This usually means that the query will be
routed to the master.
Galera Cluster variables, such as `@@wsrep_node_name`, are not resolved by the
embedded MariaDB parser. This usually means that the query will be routed to the
master.
### Limitations in client session handling
#### Limitations in client session handling
Some of the queries that client sends are routed to all backends instead
of sending them just to one of server. These queries include `USE <db
name>` and `SET autocommit=0` among many others. Readwritesplit sends a
copy of these queries to each backend server and forwards the master's
reply to the client. Below is a list of MySQL commands which are
classified as session commands :
Some of the queries that a client sends are routed to all backends instead of
just to one. These queries include `USE <db name>` and `SET autocommit=0`, among
many others. Readwritesplit sends a copy of these queries to each backend server
and forwards the master's reply to the client. Below is a list of MySQL commands
which are classified as session commands.
```
COM_INIT_DB (USE <db name> creates this)
COM_CHANGE_USER
COM_STMT_CLOSE
COM_STMT_SEND_LONG_DATA
COM_STMT_RESET
COM_STMT_PREPARE
COM_QUIT (no response, session is closed)
COM_REFRESH
COM_DEBUG
COM_PING
SQLCOM_CHANGE_DB (USE ... statements)
SQLCOM_DEALLOCATE_PREPARE
SQLCOM_PREPARE
SQLCOM_SET_OPTION
SELECT ..INTO variable|OUTFILE|DUMPFILE
SET autocommit=1|0
```
There is a possibility for misbehavior; if `USE mytable` was executed in
one of the slaves and it failed, it may be due to replication lag rather
than the fact it didn’t exist. Thus the same command may end up with
different result among backend servers. The slaves which fail to execute a
session command will be dropped from the active list of slaves for this
session to guarantee a consistent session state across all the servers
that are in use by the session.
The above-mentioned behavior can be partially controller with the
`use_sql_variables_in` configuration parameter.
There is a possibility for misbehavior. If `USE mytable` is executed in one of
the slaves and fails, it may be due to replication lag rather than the
database not existing. Thus, the same command may produce different result in
different backend servers. The slaves which fail to execute a session command
will be dropped from the active list of slaves for this session to guarantee a
consistent session state across all the servers used by the session.
The above-mentioned behavior can be partially controlled with the configuration
parameter `use_sql_variables_in`:
```
use_sql_variables_in=[master|all] (default: all)
```
Server-side session variables are called as SQL variables. If "master" is
set, SQL variables are read and written in master only. Autocommit values
and prepared statements are routed to all nodes always.
Server-side session variables are handled similar to SQL variables. If "master"
is set, SQL variables are read and written in master only. Autocommit values and
prepared statements are routed to all nodes always.
**WARNING**
If a SELECT query modifies a user variable when the `use_sql_variables_in`
parameter is set to `all`, it will not be routed and the client will receive
an error. A log message is written into the log further explaining the reason
for the error. Here is an example use of a SELECT query which modifies a user
parameter is set to `all`, it will not be routed and the client will receive an
error. A log message is written into the log further explaining the reason for
the error. Here is an example use of a SELECT query which modifies a user
variable and how MariaDB MaxScale responds to it.
```
@ -187,105 +182,54 @@ MySQL [(none)]> SELECT @id := @id + 1 FROM test.t1;
ERROR 1064 (42000): Routing query to backend failed. See the error log for further details.
```
You allow user variable modification in SELECT queries by setting the
value of `use_sql_variables_in` to `master`. This will route all queries
that use user variables to the master.
Allow user variable modification in SELECT queries by setting the value of
`use_sql_variables_in` to `master`. This will route all queries that use user
variables to the master.
#### Examples of session command limitations
If a new database "db" was created and client executes “USE db” and it is
routed to a slave before the CREATE DATABASE clause is replicated to all
slaves, there is a risk of executing a query in the wrong database. Similarly, if
any response that RWSplit sends back to the client differ from that of the
master, there is a risk for misbehavior. To prevent this, any failures in
session command execution are treated as fatal errors and all connections
by the session to that particular slave server will be closed. In
addition, the server will not used again for routing for the duration of
the session.
In a situation where a new database `db` is created, immediately after which a client executes `USE db`, it is possible that the command is routed
to a slave before the `CREATE DATABASE` clause is replicated to all slaves. In this case a query may be executed in the wrong database. Similarly, if any response
that ReadWriteSplit sends back to the client differ from that of the master,
there is a risk for misbehavior. To prevent this, any failures in session
command execution are treated as fatal errors and all connections by the session
to that particular slave server will be closed. In addition, the server will not
used again for routing for the duration of the session.
The most likely reasons are related to replication lag but it could be
possible that a slave fails to execute something because of some
non-fatal, temporary failure, while the execution of the same command
succeeds in other backends.
The most likely reasons are related to replication lag but it could be possible
that a slave fails to execute something because of some non-fatal, temporary
failure, while the execution of the same command succeeds in other backends.
## Schemarouter limitations (schemarouter)
The preparation of a prepared statement is routed to all servers. The execution
of a prepared statement is routed to the first available server or to the server
pointed by a routing hint attached to the query.
The schemarouter router currently has some limitations due to the nature of
the sharding implementation and the way the session variables are detected
and routed. Here is a list of the current limitations.
### Schemarouter limitations (schemarouter)
* Cross-database queries (e.g.
`SELECT column FROM database1.table UNION select column FROM database2.table`)
are not supported and are routed either to the first explicit database
in the query, the current database in use or to the first available
database, if none of the previous conditions are met.
The schemarouter currently has some limitations due to the nature of the
sharding implementation and the way the session variables are detected and
routed. Here is a list of the current limitations:
* Cross-database queries (e.g. `SELECT column FROM database1.table UNION select
column FROM database2.table`) are not supported and are routed either to the
first explicit database in the query, the current database in use or to the
first available database, depending on which succeeds.
* Without a default database, queries without explicit databases that do not
modify the session state will be routed the first available server. This
means that, for example when creating a new database, queries should
be done directly on the node or the router should be equipped with
the hint filter and a routing hint should be used. Queries that
modify the session state e.g. `SET autocommit=1` will be routed
to all servers regardless of the default database.
modify the session state will be routed to the first available server. This
means that, for example when creating a new database, queries should be done
directly on the node or the router should be equipped with the hint filter and a
routing hint should be used. Queries that modify the session state (e.g. `SET
autocommit=1`) will be routed to all servers regardless of the default database.
* SELECT queries that modify session variables are not currently supported
because uniform results can not be guaranteed. If such a query is
executed, the behavior of the router is undefined. To work around this
limitation the query must be executed in separate parts.
* SELECT queries that modify session variables are not currently supported because
uniform results can not be guaranteed. If such a query is executed, the behavior
of the router is undefined. To work around this limitation, the query must be
executed in separate parts.
* If a query targets a database the schemarouter hasn't mapped to a server
the query will be routed to the first available server. This possibly
returns an error about database rights instead of a missing database.
* If a query targets a database the schemarouter hasn't mapped to a server, the
query will be routed to the first available server. This possibly returns an
error about database rights instead of a missing database.
* The preparation of a prepared statement is routed to all servers. The
execution of a prepared statement is routed to the first available server or
to the server pointed by a routing hint attached to the query. As text
protocol prepared statements are relatively rare, prepared statements can't be
considered as supported in schemarouter
## Avrorouter limitations (avrorouter)
The avrorouter does not support the following data types and conversions.
* BIT
* Fields CAST from integer types to string types
The avrorouter does not do any crash recovery. This means that the avro files
need to be truncated to valid block lengths before starting the avrorouter.
# Authenticator limitations
## MySQL Authentication Related Limitations (MySQLAuth)
* MariaDB MaxScale supports authentication that uses wildcard matching in
hostnames in the mysql.user-table of the backend database. For IP-address
entries either '%' or '_'-wildcards are accepted, they should not be mixed
in the same entry. For text addresses both wildcards can be mixed.
* When wildcards are used with text-form hostnames, MariaDB MaxScale uses
the `getnameinfo()`-function (from glibc) to perform the reverse DNS lookup.
This is a slow operation, which will stall the calling thread for an
unspecified time. The lookup is only performed if the mysql.user-table
contains a text hostname with wilcards for the client. Also, do note that
the IP and hostname seen by the backend will be those of the machine running
MariaDB MaxScale, while the username and password will be those of the
original client.
* MySQL old style passwords are not supported. MySQL versions 4.1 and newer use
a new authentication protocol which does not support pre-4.1 style passwords.
* When users have different passwords based on the host from which they connect
MariaDB MaxScale is unable to determine which password it should use to connect to the
backend database. This results in failed connections and unusable usernames
in MariaDB MaxScale.
# Filter limitations
Filters are not guaranteed to receive complete MySQL packets if they are used
with the readconnroute router. This can be fixed by using the readwritesplit
router.
## Database Firewall limitations (dbfwfilter)
The Database Firewall filter does not support multi-statements. Using them
will result in an error being sent to the client.
* As text protocol prepared statements are relatively rare, prepared statements
are not supported in schemarouter.