drizzle-nginx-module - Upstream module for talking to MySQL and Drizzle directly

This module is not distributed with the Nginx source. See the installation instructions.

• Status
• Version
• Synopsis
• Description
  • Keepalive connection pool
  • Last Insert ID
• Directives
  • drizzle_server
  • drizzle_keepalive
  • drizzle_query
  • drizzle_pass
  • drizzle_connect_timeout
  • drizzle_send_query_timeout
  • drizzle_recv_cols_timeout
  • drizzle_recv_rows_timeout
  • drizzle_buffer_size
  • drizzle_module_header
  • drizzle_status
• Variables
  • $drizzle_thread_id
• Output Format
  • RDS Header Part
  • RDS Body Part
    • Columns
    • Rows
      • Row Flag
      • Fields Data
  • RDS buffer Limitations
• Status Code
• Caveats
• Trouble Shooting
• Known Issues
• Installation
• Compatibility
• Community
  • English Mailing List
  • Chinese Mailing List
• Report Bugs
• Source Repository
• Test Suite
• TODO
• Changes
• Authors
• Copyright & License
• See Also

This module is already production ready.

This document describes ngx_drizzle v0.1.11 released on 19 April 2018.

 http {
     ...

     upstream cluster {
         # simple round-robin
         drizzle_server 127.0.0.1:3306 dbname=test
              password=some_pass user=monty protocol=mysql;
         drizzle_server 127.0.0.1:1234 dbname=test2
              password=pass user=bob protocol=drizzle;
     }

     upstream backend {
         drizzle_server 127.0.0.1:3306 dbname=test
              password=some_pass user=monty protocol=mysql;
     }

     server {
         location /mysql {
             set $my_sql 'select * from cats';
             drizzle_query $my_sql;

             drizzle_pass backend;

             drizzle_connect_timeout    500ms; # default 60s
             drizzle_send_query_timeout 2s;    # default 60s
             drizzle_recv_cols_timeout  1s;    # default 60s
             drizzle_recv_rows_timeout  1s;    # default 60s
         }

         ...

         # for connection pool monitoring
         location /mysql-pool-status {
             allow 127.0.0.1;
             deny all;

             drizzle_status;
         }
     }
 }

Back to TOC

This is an nginx upstream module integrating libdrizzle into Nginx in a non-blocking and streamming way.

Essentially it provides a very efficient and flexible way for nginx internals to access MySQL, Drizzle, as well as other RDBMS's that support the Drizzle or MySQL wired protocol. Also it can serve as a direct REST interface to those RDBMS backends.

This module does not generate human-readable outputs, rather, in a binary format called Resty-DBD-Stream (RDS) designed by ourselves. You usually need other components, like rds-json-nginx-module, rds-csv-nginx-module, or lua-rds-parser, to work with this module. See Output Format for details.

Back to TOC

This module also provides a builtin per-worker connection pool mechanism for MySQL or Drizzle TCP connections.

Here's a sample configuration:

 upstream backend {
     drizzle_server 127.0.0.1:3306 dbname=test
          password=some_pass user=monty protocol=mysql;
     drizzle_keepalive max=100 mode=single overflow=reject;
 }

For now, the connection pool uses a simple LIFO algorithm to assign idle connections in the pool. That is, most recently (successfully) used connections will be reused first the next time. And new idle connections will always replace the oldest idle connections in the pool even if the pool is already full.

See the drizzle_keepalive directive for more details.

Back to TOC

If you want to get LAST_INSERT_ID, then ngx_drizzle already returns that automatically for you when you're doing a SQL insert query. Consider the following sample nginx.conf snippet:

 location /test {
     echo_location /mysql "drop table if exists foo";
     echo;
     echo_location /mysql "create table foo (id serial not null, primary key (id), val real);";
     echo;
     echo_location /mysql "insert into foo (val) values (3.1415926);";
     echo;
     echo_location /mysql "select * from foo;";
     echo;
 }

 location /mysql {
     drizzle_pass backend;
     drizzle_module_header off;
     drizzle_query $query_string;
     rds_json on;
 }

Then request GET /test gives the following outputs:

 {"errcode":0}
 {"errcode":0}
 {"errcode":0,"insert_id":1,"affected_rows":1}
 [{"id":1,"val":3.1415926}]

You can see the insert_id field (as well as the affected_rows field in the 3rd JSON response.

Back to TOC

Back to TOC

syntax: drizzle_server <host> user=<user> password=<pass> dbname=<database>

syntax: drizzle_server <host>:<port> user=<user> password=<pass> dbname=<database> protocol=<protocol> charset=<charset>

default: no

context: upstream

Directive assigns the name and the parameters of server. For the name it is possible to use a domain name, an address, with an optional port (default: 3306). If domain name resolves to several addresses, then all are used.

The following options are supported:

user=<user> MySQL/Drizzle user name <user> for login.

password=<pass> Specify mysql password <pass>for login. If you have special characters like # or spaces in your password text, then you'll have to quote the whole key-value pair with either single-quotes or double-quotes, as in

 drizzle_server 127.0.0.1:3306 user=monty "password=a b#1"
         dbname=test protocol=mysql;

dbname=<database> Specify default MySQL database <database> for the connection. Note that MySQL does allow referencing tables belonging to different databases by qualifying table names with database names in SQL queries.

protocol=<protocol> Specify which wire protocol to use, drizzle or mysql. Default to drizzle.

charset=<charset> Explicitly specify the character set for the MySQL connections. Setting this option to a non-empty value will make this module send out a set names '<charset>' query right after the mysql connection is established. If the default character encoding of the MySQL connection is already what you want, you needn't set this option because it has extra runtime cost. Here is a small example:

 drizzle_server foo.bar.com:3306 user=monty password=some_pass
                                 dbname=test protocol=mysql
                                 charset=utf8;

Please note that for the mysql server, "utf-8" is not a valid encoding name while utf8 is.

Back to TOC

syntax: drizzle_keepalive max=<size> mode=<mode>

default: drizzle_keepalive max=0 mode=single

context: upstream

Configures the keep-alive connection pool for MySQL/Drizzle connections.

The following options are supported:

max=<num> Specify the capacity of the connection pool for the current upstream block. The value must be non-zero. If set to 0, it effectively disables the connection pool. This option is default to 0.

mode=<mode> This supports two values, single and multi. The single mode means the pool does not distinguish various drizzle servers in the current upstream block while multi means the pool will merely reuse connections which have identical server host names and ports. Note that even under multi, differences between dbname or user parameters will be silently ignored. Default to single.

overflow=<action> This option specifies what to do when the connection pool is already full while new database connection is required. Either reject or ignore can be specified. In case of reject, it will reject the current request, and returns the 503 Service Unavailable error page. For ignore, this module will go on creating a new database connection.

Back to TOC

syntax: drizzle_query <sql>

default: no

context: http, server, location, location if

Specify the SQL queries sent to the Drizzle/MySQL backend.

Nginx variable interpolation is supported, but you must be careful with SQL injection attacks. You can use the set_quote_sql_str directive, for example, to quote values for SQL interpolation:

 location /cat {
     set_unescape_uri $name $arg_name;
     set_quote_sql_str $quoted_name $name;

     drizzle_query "select * from cats where name = $quoted_name";
     drizzle_pass my_backend;
 }

Back to TOC

syntax: drizzle_pass <remote>

default: no

context: location, location if

phase: content

This directive specifies the Drizzle or MySQL upstream name to be queried in the current location. The <remote> argument can be any upstream name defined with the drizzle_server directive.

Nginx variables can also be interpolated into the <remote> argument, so as to do dynamic backend routing, for example:

 upstream moon { drizzle_server ...; }

 server {
     location /cat {
         set $backend 'moon';

         drizzle_query ...;
         drizzle_pass $backend;
     }
 }

Back to TOC

syntax: drizzle_connect_time <time>

default: drizzle_connect_time 60s

context: http, server, location, location if

Specify the (total) timeout for connecting to a remote Drizzle or MySQL server.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, i.e., "second". The default setting is 60s.

Back to TOC

syntax: drizzle_send_query_timeout <time>

default: drizzle_send_query_timeout 60s

context: http, server, location, location if

Specify the (total) timeout for sending a SQL query to a remote Drizzle or MySQL server.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, ie, "second". The default setting is 60s.

Back to TOC

syntax: drizzle_recv_cols_timeout <time>

default: drizzle_recv_cols_timeout 60s

context: http, server, location, location if

Specify the (total) timeout for receiving the columns metadata of the result-set to a remote Drizzle or MySQL server.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, ie, "second". The default setting is 60s.

Back to TOC

syntax: drizzle_recv_rows_timeout <time>

default: drizzle_recv_rows_timeout 60s

context: http, server, location, location if

Specify the (total) timeout for receiving the rows data of the result-set (if any) to a remote Drizzle or MySQL server.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, ie, "second". The default setting is 60s.

Back to TOC

syntax: drizzle_buffer_size <size>

default: drizzle_buffer_size 4k/8k

context: http, server, location, location if

Specify the buffer size for drizzle outputs. Default to the page size (4k/8k). The larger the buffer, the less streammy the outputing process will be.

Back to TOC

syntax: drizzle_module_header on|off

default: drizzle_module_header on

context: http, server, location, location if

Controls whether to output the drizzle header in the response. Default on.

The drizzle module header looks like this:

X-Resty-DBD-Module: ngx_drizzle 0.1.0

Back to TOC

syntax: drizzle_status

default: no

context: location, location if

phase: content

When specified, the current Nginx location will output a status report for all the drizzle upstream servers in the virtual server of the current Nginx worker process.

The output looks like this:

worker process: 15231

upstream backend
  active connections: 0
  connection pool capacity: 10
  overflow: reject
  cached connection queue: 0
  free'd connection queue: 10
  cached connection successfully used count:
  free'd connection successfully used count: 3 0 0 0 0 0 0 0 0 0
  servers: 1
  peers: 1

upstream backend2
  active connections: 0
  connection pool capacity: 0
  servers: 1
  peers: 1

Note that, this is not the global statistics if you do have multiple Nginx worker processes configured in your nginx.conf.

Back to TOC

This module creates the following Nginx variables:

Back to TOC

This variable will be assigned a textual number of the underlying MySQL or Drizzle query thread ID when the current SQL query times out. This thread ID can be further used in a SQL kill command to cancel the timed-out query.

Here's an example:

 drizzle_connect_timeout 1s;
 drizzle_send_query_timeout 2s;
 drizzle_recv_cols_timeout 1s;
 drizzle_recv_rows_timeout 1s;

 location /query {
     drizzle_query 'select sleep(10)';
     drizzle_pass my_backend;
     rds_json on;

     more_set_headers -s 504 'X-Mysql-Tid: $drizzle_thread_id';
 }

 location /kill {
     drizzle_query "kill query $arg_tid";
     drizzle_pass my_backend;
     rds_json on;
 }

 location /main {
     content_by_lua '
         local res = ngx.location.capture("/query")
         if res.status ~= ngx.HTTP_OK then
             local tid = res.header["X-Mysql-Tid"]
             if tid and tid ~= "" then
                 ngx.location.capture("/kill", { args = {tid = tid} })
             end
             return ngx.HTTP_INTERNAL_SERVER_ERROR;
         end
         ngx.print(res.body)
     '
 }

where we make use of headers-more-nginx-module, lua-nginx-module, and rds-json-nginx-module too. When the SQL query timed out, we'll explicitly cancel it immediately. One pitfall here is that you have to add these modules in this order while building Nginx:

• lua-nginx-module
• headers-more-nginx-module
• rds-json-nginx-module

Such that, their output filters will work in the reversed order, i.e., first convert RDS to JSON, and then add our X-Mysql-Tid custom header, and finally capture the whole (subrequest) response with the Lua module. You're recommended to use the OpenResty bundle though, it ensures the module building order automatically for you.

Back to TOC

This module generates binary query results in a format that is shared among the various Nginx database driver modules like ngx_postgres. This data format is named Resty DBD Stream (RDS).

If you're a web app developer, you may be more interested in

• using rds-json-nginx-module to obtain JSON output,
• using rds-csv-nginx-module to obain Comma-Separated-Value (CSV) output,
• or using lua-rds-parser to parse the RDS data into Lua data structures.

For the HTTP response header part, the 200 OK status code should always be returned. The Content-Type header must be set to application/x-resty-dbd-stream. And the driver generating this response also sets a X-Resty-DBD header. For instance, this module adds the following output header:

X-Resty-DBD-Module: drizzle 0.1.0

where 0.1.0 is this module's own version number. This X-Resty-DBD-Module header is optional though.

Below is the HTTP response body format (version 0.0.3):

Back to TOC

The RDS Header Part consists of the following fields:

uint8_t endian type (1 means big-endian and little endian otherwise)

uint32_t format version (v1.2.3 is represented as 1002003 in decimal)

uint8_t result type (0 means normal SQL result type, fixed for now)

uint16_t standard error code

uint16_t driver-specific error code

uint16_t driver-specific error string length

**u_char *** driver-specific error string data

uint64_t database rows affected

uint64_t insert id (if none, 0)

uint16_t column count

Back to TOC

When the column count field in the RDS Header Part is zero, then the whole RDS Body Part is omitted.

The RDS Body Part consists of two sections, Columns and Rows.

The columns part consists of zero or more column data. The number of columns is determined by column count field in RDS Header Part.

Each column consists of the following fields

uint16_t non-zero value for standard column type code and for the column list terminator and zero otherwise.

uint16_t driver-specific column type code

uint16_t column name length

**u_char *** column name data

The rows part consists of zero or more row data, terminated by a 8-bit zero.

Each row data consists of a Row Flag and an optional Fields Data part.

uint8_t valid row (1 means valid, and 0 means the row list terminator)

The Fields Data consists zero or more fields of data. The field count is predetermined by the column number</code) specified in RDS Header Part.

     mysql --protocol=tcp -u user --password=password -h foo.bar.com dbname

Note that the `--protocol=tcp` option is required here, or your MySQL client may use Unix Domain Socket to connect to your MySQL server.

     tar xzvf drizzle7-2011.07.21.tar.gz
     cd drizzle7-2011.07.21/
     ./configure --without-server
     make libdrizzle-1.0
     make install-libdrizzle-1.0

Ensure that you have the `python` command point to a `python2` interpreter. It's known that on recent : Arch Linux distribution, `python` is linked to `python3` by default, and while running `make libdrizzle-1.0` will yield the error

     File "config/pandora-plugin", line 185
         print "Dependency loop detected with %s" % plugin['name']
                                                  ^
     SyntaxError: invalid syntax
     make: *** [.plugin.scan] Error 1

You can fix this by pointing `python` to `python2`.

     wget 'http://nginx.org/download/nginx-1.13.6.tar.gz'
     tar -xzvf nginx-1.13.6.tar.gz
     cd nginx-1.13.6/

     # if you have installed libdrizzle to the prefix /opt/drizzle, then
     # specify the following environments:
     # export LIBDRIZZLE_INC=/opt/drizzle/include/libdrizzle-1.0
     # export LIBDRIZZLE_LIB=/opt/drizzle/lib

     # Here we assume you would install you nginx under /opt/nginx/.
     ./configure --prefix=/opt/nginx \
                 --add-module=/path/to/drizzle-nginx-module

     make -j2
     make install

 $ PATH=/path/to/your/nginx-with-echo-module:$PATH prove -r t