1. This is the 40th article of the public No.
2. Go without knowing the language, reading this article will not be obstacle
3. Cover images from Mael BALLAND on Unsplash
What is the pool of technology
database/sql 包
Design Philosophy
Minimalist Interface
Call relationship
Connection Pool Design
sql.DB key attributes of the object
establish connection
Connection release
Clean-up connection
to sum up
What is the pool of technology
Pooling technology ( Pool) is a very common programming techniques, when requesting a large amount significantly optimize application performance and reduce system resource overhead frequently even built. Our daily work in common with database connection pooling, thread pooling, object pooling, etc., which are the characteristics of the "expensive" and "time-consuming" maintenance of resources in a particular "pool" in the provisions of its minimum number of connections the maximum number of connections, blocking queue configuration, etc., to facilitate unified management and reuse, also typically comes with a number of exploration alive mechanism, mandatory recycling, monitoring a kind of supporting functions.
database/sql 包
Design Philosophy
In the Go language database operations required by the standard library database/sql package, its upper application provides a standard API user interface, on the exposed lower driver driving a simple interface, and internal connection pool management realized. This means that drivers of different databases can easily drive to achieve these interfaces, but no longer need to care about the details of the connection pool, just based on a single connection.
Minimalist Interface
It exposed the external interface is simple and easy to understand, which will help third parties Driver to achieve, functional interface, including Driver registration, Conn, Stmt, Tx, Rowsthe result set, and we by Conn and Stmt these two interfaces to experience the exquisite interface design (which corresponds to two interfaces Java is Connection and Statement the interface, but Go simpler)
I believe that even if you have not learned Go the language, alone your Java knowledge, you can effortlessly understand the meaning of these interfaces above, these layers are exposed to the drive interface is very simple, so that the driver can easily to achieve.
Call relationship
The entire database/sql call relationship-driven interface is very clear and simple terms by drivers first Open method to get a new Conn, and then call Conn the Prepare method, passing in SQL the statement is the statement Stmt, the last call Stmt the Exec method returns a result parameter passing, query empathy , but returns the result set of data rows.
Connection Pool Design
sql.DB key attributes of the object
Go When the operation of the database language, we use the first sql.Open method to return a specific sql.DB target, the film following code db :
sql.DB The object that is the entrance of our access to the database, we look at key attributes inside it, were associated with the connection pool design
establish connection
In fact, the connection is not in sql.Open the return db on the establishment of an object, which just opened a reception step up a connection request channel, even the actual construction of concrete steps to be performed until SQL when the statement will be. Here is how we achieve by telling about some examples of logical connection pool connection is established how.
This section describes the principle source does not stick too much, it becomes a source resolved, the students do not understand the language Go unfriendly, mainly hoping to convey some ideas connection pool design.
In the database/sql upper applications exposed in the user interface, it is more commonly used Exec and Query, the former is used to perform the write SQL, which can be used for reading SQL. But regardless of which method to go, you will be called to build even logical db.conn method, and even comes with built context and strategies to build even two parameters.
Which even built into policies cachedOrNewConn and alwaysNewConn. From the former preferentially freeConn removed connection idle connection, or create a new one; the latter always walk logic new connection.
Using cachedOrNewConn up a connection in a logical strategy, first determines whether there is an idle connection, if there is an idle connection is first removed, followed by determining whether the connection needs to be recovered expired, and if not expired can be used normally goes to a subsequent logic. If no free connection is determined that the number of connections is not already reached the maximum, if no new connection, otherwise it must wait for an available blocking connection request.
If desired new connection, is called the bottom Driver connector implemented Connect interface, which is part of a respective database Driver itself to achieve.
Connection release
After the connection is a need to return to the pool after use, which is the database connection pool to achieve the more important logical, usually accompanied on the reliability of the detection connection, if the connection is closed, then returned to the connection pool should not continue, and Create a new connection should be replaced.
In Java the Druid connection pool will testOnReturn or testOnBorrow an option, it is checked for validity at the time of return connection or get a connection, but the open nature will be extended on both connection occupied time, lose some performance. Go Language to achieve this function is relatively simple, and there is no specific validity detection mechanism, but a direct connection according to the supplied err information, if it is ErrBadConn abnormal shut down and send a new signal.
Clean-up connection
database/sql It provides three key parameters related to the connection pool under the package, respectively maxIdle, maxOpen and maxLifeTime.
Meaning three parameters is easy to understand, if you want to understand, recommended reading Configuring sql.DB for Better Performance
MySQL Side will be forced kill out of idle connection time (8h), Go language provides maxLifeTime the maximum time set the connection options are multiplexed, attention is not idle connection time, but to establish a connection from this point in time will be recovered, so as to ensure connection active.
This cleaning mechanism is through an asynchronous task to do, each key is logically one second traverse checking freeConn an idle connection is determined whether multiplexing period exceeds the maximum, beyond the connecting join Closing array is subsequently Close.
to sum up
Recent work content is based on the go-sql-driver realization of a custom support a separate read and write and highly available driver, during the research and learning to feel the Go language database/sql clear and concise package, although it is in the realization of some of the features of simple partial or no, but still covered the main functions and features most of the database connection pool, so I think to use it to start a discussion is a good choice.
