Weidian Chen Jian: The road to mobile terminal optimization from the perspective of operation and maintenance engineers

This article is the on-site dry goods of WOT2016 Internet Operation and Developer Conference. The new session with the theme of WOT2016 Enterprise Security Technology Summit will be held at JW Marriott Hotel in Beijing Pearl River Delta from June 24th to 25th, 2016!

Chen Jian is an operation and maintenance engineer at Weidian. He is mainly engaged in mobile operation and maintenance and operation and maintenance automation.

[[167110]]

Since 2003, Weidian has been able to provide users with decentralized services through mobile phones, completing most of the functions similar to Taobao or third-party stores, and can share store information with friends or WeChat Moments at any time. Users can also complete operations such as store location marking and business license application through the PC version of Weidian. At the same time, there are also functions such as pocket shopping, buyer's version of global shopping, today's half price, and overseas version of U-shop.

Compared with other speeches, Chen Jian's topic is more about operation and maintenance, mainly summarized from the perspective of technology. He explained from the perspectives of basic services, automation and monitoring:

Basic Services

In terms of basic services, Chen Jian explained how to build and optimize the Weidian mobile terminal from the aspects of NGINX, PHP, DNS, TOMCAT, etc.

NGINX . NGINX has many services, through which you can learn how to obtain the real IP through remote_addr in NAT mode, add request link information to the Header, optimize the Location rewrite sequence, redirect abandoned domain names more elegantly, and implement global cross-domain support, etc. Most domestic services are built under NGINX.

PHP. PHP is the core of the backend programming language, and all external service API interfaces are written in PHP. The above are the commonly used parameters of PHP.

Through real-time analysis, we can see some monitoring data, through which we can find out which interfaces may have problems, whether caused by development or downstream services, and find some slow requests in the first place.

DNS , like some traditional Internet, now gives priority to using DNS for intranet resolution, but a layer of DNSmasq is hung in front of DNS. Let me talk about the configuration first. The default time is changed to 1 second, and the minimum value is also 1 second. A total of one local plus three primary DNS are configured. When doing stress testing on the original DNS, it was obvious that many interfaces would be relatively poor. Later, after the transformation, the intranet DNS is much more robust than before, and the pressure on the DNS master station is also reduced a lot. This service will be configured by default when installing the machine, and it will be started through the program. Once it hangs, it will automatically pull up, but so far, there has been no active hang.

TOMCAT , each company has different tuning according to its own business situation. In the TOMCAT area, we mainly did the following: We customized the port and controlled its range between 2200-2300. We allow no more than ten TOMCATs on a single machine. Once a program has a problem, it will not affect other services. Now the new architecture is all run on a single machine with a single strength and a single TOMCAT, using programs similar to KM. First, get the IP address of the local machine. This value is used for JMX monitoring. We will add up the monitoring rights of this machine through dynamic scripting, including taking out TOMDIR. This variable will be used later, mainly to delete the cache key to ensure that every time TOMCAT is restarted, it is a completely clean environment.

Network optimization. First of all, the core network is BGP, and there are China Unicom, China Telecom, and China Mobile dedicated lines. Many people may wonder why we need to use other lines after BGP. Cost considerations are one aspect, but more importantly, BGP is not as good as we imagined, especially when it comes to inter-network settlement, which has a greater impact on the network in some places. When there was no automatic domain name link splitting, there might be several complaints every day that the WeChat store could not be opened, or there was a problem with a certain activity detail page. Later, all the links were completely split, and then all the links in multiple computer rooms were opened. It has been about three years since the development of WeChat store. Our scale may still be much smaller than that of most large Internet companies. At present, this structure is a relatively reasonable solution for us.

Chen Jian explained the problems encountered in the WeChat store construction process and how they were solved subsequently from the aspects of HTTPDNS, HTTPS content interception, image optimization and automation.

Approval mechanism. For developers, the daily online time is from 9 am to 6 pm on weekdays. Code can be submitted during this period. Some projects, such as core services, will require full approval.

Monitoring. Monitoring is very important for operation and maintenance. I will not talk about basic monitoring, but mainly talk about the application-level monitoring currently being done.

This is a minute-level monitoring. In fact, each graph has two lines. On the mobile monitoring page, we often communicate with users and want to know some basic information, such as your IP address, DNS address, and the CDN node you are currently accessing. This information is actually very difficult to obtain. Later, we found a way to develop a monitoring system like this. Users only need to access this address in WeChat or a browser to obtain this basic information.

Big data monitoring system. Each of our businesses is composed of interfaces and all ULs. After selecting a UL, you can know some basic information about the UL, including error rates, QUVS, etc. This is what we are doing in the big data field.

Lecture video: http://edu..com/lesson/id-100754.html

【Lecturer Profile】

[[167112]]

Lecturer Profile: Chen Jian, Weidian operation and maintenance expert, was one of the first batch of people to join Weidian. He has witnessed the development history of Weidian and has been working hard for the "stability", "efficiency" and "security" of Weidian. He is very proficient in networks, operating systems and applications, and is currently focusing on the continuous optimization of Weidian's access.