Large Web site Technical architecture Read Note 5

Large Web site Technical architecture Read Note 5

This time, I mainly read the book's analysis of typical failure cases on large Web sites and the architect's leadership in architecture architects.

General fault phenomenon, due to the release of an application, the database load is high, far beyond the normal level, continuous alarm. The main reason is to check the database, found that the alarm is caused by a SQL, this is a simple indexed data query, should not raise the alarm. Continue to check, found that the SQL execution frequency is very high, far beyond the normal level, trace the SQL, found by the site application calls, the home page is the most frequently visited pages, this SQL is called by the homepage, it is frequently executed. Lessons learned, home page should not access the database, home page needs data can be obtained from the cache server or search engine server. The home page is preferably static.

Symptom: An application server is uncertain because of the response time-out and alarm, but soon again time out, return to normal, so repeated, so that operations personnel very distressed. Cause analysis, a single object in the program uses this in many places, because the this object can only have one, all concurrent requests are queued to obtain the only lock. In general, there are some simple operations, get the lock, quickly complete the operation, release the lock, will not cause the thread queue, but a remote call operation is also added this, the operation is only occasionally executed, but each execution takes a long time to complete, this time lock is occupied, All user threads Wait, the response times out, the lock is released when the operation completes, and the other threads execute quickly, and the timeout is lifted.

Symptom: After some apps are released, the server crashes immediately, causes analysis, the application Web environment uses patterns, user requests are forwarded via Apache, and Apache and JBoss launch simultaneously at release time, since JBoss startup requires many applications to be loaded and initialized, it takes a long time to As a result, JBoss has not fully started, Apache has started to accept user requests, a large number of requests blocking process, resulting in a crash, in addition to this, the site has a similar scenario, all need to prepare the background service, the foreground application can be started, otherwise it will lead to failure, this situation was internal personnel, A software technology predecessors once said, software design has two styles, one is the software design is very complex, so that its defects are not so obvious, one is the software design is very simple, with its no obvious defects, on the author observes that the two styles of software engineers are many, just in the Internet company, the latter a more some, Because even if the defect is not obvious in the rapid development of the site impact, it will soon appear alone, so that its author tired of coping, a bad, eat a loss, learn a good, later design software will be designed simple, if the problem can be found quickly, to solve is relatively easy.

We must firmly believe that a group of excellent people do a thing they love, will be able to achieve success, no matter how tortuous the process, no matter how strange the outsider seems to be unreliable. Find a goal that is worth working together and create a work environment where you can maximize your self-worth.

Large Web site Technical architecture Read Note 5