How do technicians send notifications?

tips： The documentation will continue to improve , Readers' comments are also welcome to supplement and perfect

Preface

This refers to the notification sent by the developer , For example, the following situations ：

1. The execution log of the scheduled task is sent to the developer for reference .
   For example, the task of regularly sending notifications to certain people , You can send the number of people to the developer each time , When the number of people is abnormal , It can be perceived and handled in time .
2. Whether the data logic in the offline task patrol business is abnormal , The patrol inspection results are sent to the developer for reference .
3. Various exceptions in the process of key business execution , Send to development

One 、 The most important principle - Effective communication

1. The information is valid

Of the notifications sent , Don't carry useless information .
If a notice you won't read at all , Then don't send .

2. People can handle

The amount of information a person can process is limited , therefore ：

1. The amount of information in a single notice cannot exceed the processing capacity of the person .
   For example, send all exceptions during task execution to developers , So many exceptions , Developers can see it 、 Can you handle it ？ If not , Then the part of the information that cannot be processed is invalid information .
2. Notification frequency should not be too high
   This is a common problem of email notification in large companies . I like to send emails , As soon as you open your mailbox, there are hundreds of emails , It's all done , There's nothing else to do today , So I can only choose to ignore .

3. Confirm delivery

Make sure that after the notification is sent , Relevant users will receive and view .
For example, some companies , I like sending emails without restraint , Everyone opens his mailbox every day with hundreds of emails , People are not looking at email very much . Then if the notification is sent by mail , In fact, it's no different from "no way" .

Two 、 Notification content elements

1. Time

• Each record in the notification , You need to add a time prefix
  In this way, when you have questions about the contents of the notice , It can accurately locate the specific time of occurrence , And query relevant logs in which time period 、 Monitoring, etc .
• Business logic execution time
  Monitoring notifications for complete business logic , For example, scheduled tasks 、 Interface implementation, etc , You need to know how long it takes for the entire logic to complete , And the execution time of each internal sub logic module （ For example, how much time does the third-party interface call take ）.

2. Macro index statistics and specific information sampling are combined

The statistics of macro indicators is to control the overall implementation of tasks .
Specific information sampling , When we want to verify and troubleshoot , Unable to process all log records , But we can't ignore it , So take a sample , Verify and check the sampling records .

3、 ... and 、 Other considerations

1. Send a copy to yourself , Confirm that the notification logic is executing

No matter what the notice is , Whether it's successful or not , It is better to send a copy to yourself , The main purpose is to confirm that the notification logic is working properly , And the contents of the notice are correct .
If the notification frequency is high , In the logic of sending a notice to yourself , Add a frequency limit .

Four 、 Best practices

1. Timing task ： to 100w Users send private messages

Notice template ：

[2021-12-10 01:54:56]  Timing task - Private letter notification sending - Start 
 Total number of users to send ：1,000,000
     success ：50,000
     Failure ：50,000
 Send the successful user Id sampling ：[1,2,3,4,5]
 Abnormal sampling ：[" user (6) Invalid ", " user (7) Cannot receive private messages ", " Private message interface call exception ：null"]
[2021-12-10 01:54:56]  end 

Element analysis ：

1. Start and end time
2. Macro index statistics ： The conclusion that we are most concerned about
3. Data sampling ：
   1. Send sample successfully ： We can go through the process to verify whether the user has actually sent successfully , To test logical correctness .
   2. Send failed exception sampling ： Look at the sampling information , To fix specific exceptions
4. Format large numbers , Easier to read
   Macro index data , Use the three digit separated formatting method , It is convenient to view the order of magnitude of specific values more quickly . If it is an indicator of more than one billion , At first sight , How many 0 It's just a hassle , I have to worry about the wrong number .