Shulou(Shulou.com)05/31 Report--

How to speed up the rapid iteration of enterprise IT services with the help of configuration center ACM? aiming at this problem, this article introduces the corresponding analysis and solution in detail, hoping to help more partners who want to solve this problem to find a more simple and feasible method.

Introduction to the basic form of configuration Center

With the configuration center, users or administrators no longer need to log in to the configuration center or the operating system under the configuration center to modify files. Instead, go to the configuration center platform to modify the corresponding configuration, then the configuration will listen to different applications behind, when the configuration or application starts to get the configuration, complete the task distribution through the interaction between the configuration center and the application. In this way, the efficiency of modifying configuration is greatly improved and the process is simplified.

Introduction to ACM

Application configuration Management (Application Configuration Management, referred to as ACM), ACM, which is extended by the Ali middleware configuration center tool Diamond, is the most widely used middleware within Ali, and officially became an Aliyun product in October 2017.

ACM comes from Diamond, the management tool of Ali configuration Center, so what is the situation of Diamond inside Ali? It is mainly about a few points. The first point Diamond is probably the most widely used intermediary within Ali, with a coverage of about 80%, mainly because it uses a lot of scenarios. As a configuration center tool, there is an excellent performance guarantee within Ali, and the internal commitment of SLA is within 1 second, which provides a capability guarantee for many upper-level businesses. When there is a failure in the computer room, the configuration center can ensure that a lot of workloads can be migrated within 1 second. There are many other various scenarios, for example, through very convenient operation and maintenance capabilities can help many other businesses of the group to do similar to current limit and downgrade. There are also scenarios to do some algorithms, and to do multi-environment management, and there are many application scenarios, so this is the current situation of the use of configuration center management in Ali.

Externally, what is the historical trend of the configuration center in the industry in the same period? today, the configuration center has completed a historical development path from the original non-tool to having tools and becoming a formal product on the public cloud. From the development process of the industry, the configuration center has completed a development process from no tools to tools and then to formal production.

ACM Technology explains the basic functions of configuration Center in detail

After having the configuration center, the user does not need to log in to the machine to change the configuration file. Instead, he publishes a configuration on the configuration management console, and the corresponding configuration is written to the server in the configuration center. The service center will push different configurations to the following applications. So there are three basic functions, the first is to release the configuration, the second is to obtain the configuration, and the third is to monitor the change of the dynamic configuration.

Infrastructure construction

The server of ACM is mainly composed of three parts. The first is the address server, which provides ACM (Diamond) service discovery function, which realizes the transparency of cluster expansion, offline and other operation and maintenance work to the application side. The second is the ACM (diamond)-server component, which is the core of ACM services. This component provides persistent management of configuration and dynamic configuration push service to the application side through Http protocol. In implementation, the ACM (diamond)-server component is a Web service cluster, and the cluster nodes realize the incremental synchronization of data changes through a specific mechanism, which is a decentralized structure. The third is Mysql storage, which is the basis for persistent configuration management of ACM (Diamond) services. In Taobao production environment, the deployment mode of "one master and two backups" is used to ensure the security of data.

To sum up, the design of ACM uses clusters instead of centers to achieve no single point, so as to improve the reliability of the system; using Http stateless protocol to expose services, so that the system logic is simple, not easy to make mistakes.

The principle of configuration release and acquisition based on Cache

The persistent configuration of ACM (Diamond) is saved in mysql, and each node of the additional server saves all the configuration data as cache in the form of a local disk file. After introducing the cache file of the server configuration data, a read request of the configuration data is actually a Http request of a static file. With the optimization of 0 copy, the useless memory copy of configuration data between application state and kernel state is reduced, and the efficiency of data transmission is greatly improved.

With the introduction of cache mechanism, ACM (Diamond) can be defined as a distributed cache system in implementation. In the aspect of CAP, ACM (Diamond) chooses to sacrifice the strong consistency of data to provide the final consistency, and ensures the availability and partition tolerance of the system. It is also proved by practice that providing the final consistent configuration and concurrent read and write can meet almost all configuration center application scenarios.

Configure snooping

The effect of configuration monitoring is a little different from that of publishing and obtaining. When a configuration changes, how to push the change to the client through a similar message, the client can know the change in time and take some corresponding actions internally. What we do is a way similar to long polling. The client will periodically initiate a 30-second persistent connection when initiating the monitoring configuration, and the persistent connection will bring it to the configuration it needs. If there is no configuration change in these 30 seconds, a null value will be returned within 30 seconds, and another connection will be initiated in the next 30 seconds. So if any change is found within these 30 seconds, the server will timely send the changed configuration through the persistent connection, and then the connection will return to the client, and the client can get the change value. This push-pull strategy achieves the balance between long connection and short connection, enables the server not to pay too much attention to connection management, and achieves the real-time performance of information push similar to TCP long connection.

Main application scenarios of ACM

Traditional configuration files can not well solve the configuration management requirements of distributed systems. Although ACM is a tool, it has a lot of energy, because it can extend many scenarios through the configuration center itself. ACM can well solve the corresponding requirements for configuration security management, release environment management, package release management, dynamic push of business scenarios, dynamic algorithm push and so on.

Scenario 1 configuration management under microservice application architecture

In the process of application release of traditional architecture, modifying an application configuration requires repackaging and publishing the whole application, which is very tedious and error-prone. In the ACM-based micro-service scenario, the important configuration information of the application is published to ACM. New configuration releases do not rely on configuration packaging. After the new version of the configuration is released, all applications will take effect immediately, as shown in the figure.

The advantage of using ACM as the configuration center for micro-services is that all configurations are centralized and configuration management becomes more convenient in the case of many applications. All configurations are independent of version release, making configuration changes more flexible. ACM naturally supports grayscale publishing and rollback, making it more secure for configuration changes to be released in a micro-service architecture.

Scenario 2 Service governance under distributed architecture

How to prevent the avalanche effect caused by performance overload, we must do some dynamic flow control. The traditional method is to change the configuration file, and after you have the configuration center, you can also present the corresponding configuration changes in the configuration center, so when there is a change in the flow control center, you only need to change the flow control threshold in the configuration center. The threshold will be dynamically pushed to all the following ordering threshold clients, so a corresponding flow control will be achieved.

The advantage of using ACM for service governance in distributed architecture is good performance, and there is almost no impact on performance by using configuration push to monitor service governance information. The relevant service governance information can be pushed in seconds, and the response time is fast. When the current limit is degraded, the state can also be restored by a second configuration rollback.

Scenario 3 Application Business scenario dynamic push

Nowadays, e-commerce will always have some operational activities. For example, International Children's Day will promote some products related to children during the corresponding period of time. The most conventional method that you will think of right away is to make a new release and push a batch of applications at that time, so that the application will change accordingly. However, after using the configuration center, the process will be greatly simplified. Specifically, write the corresponding scenario to the configuration center. In the business code displayed on the server, it is actually subscribing to a section of code in the configuration center. If the code needs to make a change in the configuration center when the scenario changes, all the business codes will be pushed to the server in time, then the corresponding scenario will change. In this way, the frequency of release is greatly reduced and the agility of the business is improved.

Scenario 4 big data real-time calculation algorithm adjustment

This is a very typical example of monitoring in the industry. When we do performance monitoring, the data of operating systems and applications will be summarized into stream computing through message pairs. How to do real-time alarm when monitoring? There are many distributed nodes when doing calculation, and all nodes need to be notified when the alarm threshold changes. In this block Ali is also done through the configuration center, the application of dynamic configuration of computing parameters, dynamic effective, effective time block, low performance impact.

Scenario 5 remote and multi-live scenarios under the enterprise Internet architecture

There are many scenarios in which ACM will be used in multi-live tasks in different places. Here is only a simple example. Within Alibaba, the core algorithm of disaster recovery and multi-activity architecture, ID shards and corresponding routing rules all use ACM to dynamically push. Among them, the corresponding client and server, such as RPC,MQ,DB, are implanted with routing paths. When a disaster recovery drill or a real disaster occurs, the administrator only needs dynamic push rules, which will affect all architectural components. The infrastructure is decoupled from the disaster recovery logic, and the specific routing logic is determined by the switching of disaster recovery rules. It takes effect quickly, and theoretically, the switching rules for disaster recovery can be pushed to 100,000-level machines in seconds.

Demo

During the transformation of the user center, it is found that users have two requirements. The first point is to consider the efficiency of operation and maintenance, changing the level of configuration changes from the level of a few hours before to the level of minutes to seconds, resulting in a great improvement in efficiency and a sharp reduction in operation and maintenance costs. The second point is the demand for safety.

Why does the configuration Center not need these sensitive information in its code after using it?

After using the configuration center, extract all the data of these businesses and put them in the configuration center. After release, all these images are the same. When you need to change the configuration, you only need to go to the console to modify the release.

If you skip the release section on ACM, how can you ensure that the configuration can be rolled back quickly and how can you record the history of release?

ACM has made a feature of the historical version in this respect. When users need to correct their mistakes, they can check the previously released historical version, find the corresponding version and roll back immediately to return to the previous version.

This is the answer to the question on how to speed up the rapid iteration of enterprise IT services with the help of the configuration center ACM. I hope the above content can be of some help to you. If you still have a lot of doubts to be solved, you can follow the industry information channel for more related knowledge.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.