What can testers do when there is an online bug?

　　 When it comes to online problems , Many test Xiaobai either ” Principle “ Fear , Or as simple as , I don't know how to start .

　　 This article , Let me refine this common interview question , Talk to everyone about online problems , As a tester , There should be some ideas .

　　 First , Directly give the three-step formula of wanjinyou ：

　　 First step , Preliminary investigation , Fast business recovery ;

　　 The second step , Find the root cause of the problem , Thorough solution ;

　　 The third step , Team sharing , Avoid similar problems .

　　 The wording in these three steps , Very important . Especially the first point —— Preliminary investigation , Fast business recovery .

　　 Problems arise , Don't blindly position yourself as soon as you come , Deal with sudden problems with the idea of never giving up when you can't find the root cause , Is not desirable .

　　 There's a problem online , The most important thing is to restore business quickly .

　　 You can check first CPU、 Memory 、 The Internet IO、 disk IO etc. , See if there is any obvious jitter . such as CPU Too high , You can try to restart .

　　 Then look at the call , Decision is a system dependent problem , Or the problem of its own system . If it's a system dependent problem , If there is a downgrade scheme, it is preferred to downgrade , If there is no degradation scheme, immediately contact the person in charge of the dependent system to solve it .

　　

　　 If it is a system problem , Give priority to database problems . If there is slow query , First kill fall , Restart the database ; If the number of visits is insufficient , Expand capacity or limit current first . Check again Full GC, If Full GC Too much , Restart the service first , Re pass DUMP Memory find object , Fix and go online .

　　 As mentioned above , Both mentioned restarting the service —— Everyone says it's good to restart the Dharma , Because it's really fragrant .

　　 occasionally , Restart is a way to quickly restore business . But this way , Treat the symptoms, not the root cause .

　　 The root cause is after the rapid resumption of business , Location problem , To solve it completely ; To make a summary and Analysis , Avoid similar problems .

　　 such as , Just mentioned , If the database is slow to query , Want to restore business quickly , Sure kill Slow down query , Restart the database . here , Although the business has resumed , But it's a short recovery , You have to keep positioning . You can locate the problem because the newly added table index is not enough , Then you have to index it right away , And hold a meeting afterwards or make a group summary , Let's talk about the design of Library tables , Talk about the online scheme and so on , Avoid similar problems .

　　 Actually , In production environment , Cause a large area of failure , There are three general types of system unavailability ： Dependent system failure 、 Database failure 、 Program problems cause insufficient memory and cause Full GC.

　　 Please remember and recite these three categories , Absolutely practical !!!

　　 More details , Database slow check 、 Deadlock 、 Insufficient connections ,redis Have big key,Full GC Too much , Threads DUMP, Memory DUMP,MQ Consumption is overstocked , Are common online questions , It can also be said that , Is most of the problem .

　　 These content , Everyone can open a special topic to talk about , Therefore, this article will not expand .

　　 In large and medium-sized companies , These are general knowledge points . There is an online problem , Experience and practice several times , Naturally understand .

　　 As a test , You can't go deep into them , But you have to know them , Or say , You can use them as a timetable for your advanced promotion , Learn one by one .

　　 Last , I hope no one in the world will encounter online BUG.