MySQL/OTP + hnc provides connection pooling for MySQL/OTP using hnc. It contains convenience functions for executing SQL queries on a connection in a pool and lets you choose between two methods for creating and managing connection pools:
- Use it as a library that helps you supervise your own MySQL connection pools.
- Use it as an application that manages its own supervisor for connection pools.
Use mysql_hnc:child_spec/3
to get a child spec for a pool that you can use
in your own supervisor.
%% my own supervisor
init([]) ->
PoolOptions = #{size => {5, 10}},
MySqlOptions = [{user, "aladdin"}, {password, "sesame"}, {database, "test"},
{prepare, [{foo, "SELECT * FROM foo WHERE id=?"}]}],
ChildSpecs = [
%% MySQL pools
mysql_hnc:child_spec(pool1, PoolOptions, MySqlOptions),
%% other workers...
{some_other_worker, {some_other_worker, start_link, []},
permanent, 10, worker, [some_other_worker]}
],
{ok, #{strategy => one_for_one}, ChildSpecs}}.
This approach requires you to start the application mysql_hnc
. Typically
this is done by adding {applications, [mysql_hnc]}
to your .app.src
file and then relying on your favourite release tool for the rest.
Pools can be added at run-time using mysql_hnc:add_pool/3
.
Pools can also be created at start-up by defining configuration parameters for
mysql_hnc
. The name of each configuration parameter is the pool name and
the value is a pair on the form {PoolOptions, MySqlOptions}
.
Example:
Start your Erlang node with erl -config mypools.config
where mypools.config
is a file with the following contents:
[
{mysql_hnc, [
{pool1, {#{size => {5, 10}},
[{user, "aladdin"}, {password, "sesame"}, {database, "test"},
{prepare, [{foo, "SELECT * FROM foo WHERE id=?"}]}]}}
]}].
The most commonly used MySQL functions are available with wrappers in
mysql_hnc
.
1> mysql_hnc:query(pool1, "SELECT * FROM foo WHERE id=?", [42]).
{ok,[<<"id">>,<<"bar">>],[[42,<<"baz">>]]}
2> mysql_hnc:execute(pool1, foo, [42]).
{ok,[<<"id">>,<<"bar">>],[[42,<<"baz">>]]}
For transactions, the connection pid is passed to the transaction fun as the first parameter.
3> mysql_hnc:transaction(pool1, fun (Pid) ->
ok = mysql:query(Pid, "INSERT INTO foo VALUES (?, ?)", [1, <<"banana">>]),
ok = mysql:query(Pid, "INSERT INTO foo VALUES (?, ?)", [2, <<"kiwi">>]),
hello
end).
{atomic, hello}
Sometimes you need to checkout a connection to execute multiple queries on it,
without wrapping it in an SQL transaction. For this purpose you can use either
a pair of calls to checkout/1
and checkin/1
or a call to with/2
with a
fun as in this example:
4> mysql_hnc:with(pool1, fun (Pid) ->
{ok, _, [[OldTz]]} = mysql:query(Pid, "SELECT @@time_zone"),
ok = mysql:query(Pid, "SET time_zone = '+00:00'"),
%% Do some stuff in the UTC time zone...
ok = mysql:query(Pid, "SET time_zone = ?", [OldTz])
end).
ok
The function checkout/1
does not return a MySQL connection directly but a
connection identifier (actually, a hnc
worker identifier). You have to
unpack the real connection with a call to get_connection/1
first.
Accordingly, the function checkin/1
does not expect a MySQL connection but
the connection identifier from which it was unpacked.
Connections to a MySQL server will be established and kept open in the pool.
Using the mysql_hnc
functions like query
, execute
, transaction
etc
will use a random connection from the pool to execute a query, ie it is unlikely
that a mysql_hnc:query
or similar call will be executed using the same connection
as the ones before.
If you are using session data like user variables, temporary tables etc, ie things that are bound to the connection that creates them and gets destroyed when it is closed, this means two things:
-
Any session data you created by a
mysql_hnc:query
(or similar) call may not exist or have different values when you usemysql_hnc:query
again, as it will likely be executed using a different connection. This behavior cannot be changed, you have to keep it in mind when writing your code. -
Any session data you created by a
mysql_hnc:query
(or similar) call will still exist when another process executes a query and is given this connection, ie the session is "dirty". Another possible scenario is that one process changed the connection to use another user, in which case it will be logged in as that user when another process gets this connection. This can be circumvented by using the on_return callback option ofhnc
(it will be called whenever a worker returns to the pool). You may use this callback to either callmysql:reset_connection
ormysql:change_user
in order to automatically reset the connection to a clean state when the connection is returned to the pool.
1> OnReturn = fun (Conn) -> mysql:change_user(Conn, "user0", "password0") end.
#Fun<erl_eval.44.97283095>
2> mysql_hnc:add_pool(pool1, #{on_return => {OnReturn, 5000}}, [{user, "user0"}, {password, "password0"}]).
{ok,<0.116.0>}
Using erlang.mk, put this in your Makefile
:
DEPS = mysql_hnc
dep_mysql_hnc = git https://github.com/mysql-otp/mysql-otp-hnc 0.3.0
Using rebar, put this in your rebar.config
:
{deps, [
{mysql_hnc, ".*", {git, "https://github.com/mysql-otp/mysql-otp-hnc",
{tag, "0.3.0"}}}
]}.
GNU Lesser General Public License (LGPL) version 3 or any later version. Since the LGPL is a set of additional permissions on top of the GPL, both license texts are included in the files COPYING.LESSER and COPYING respectively.