At present, with the increase in the electricity supplier festivals (6.18, double eleven, double XII), the new platform tends to pull the frequent big promotion activities are increasingly common. As a business platform, there will be one, or even several times the flow rate of the "final exam" every year. As an important node database system, its stability and performance particularly important, to protect the database is a big challenge. Electricity supplier big promotion, this war did not smoke a lot of people have experience, not discussed here. Now, we confront a subject directly - Database how to actively respond, to protect big promotion activities. Explain this topic decomposition into three parts: The first part of the preparatory work; the second part, the promotion of large row; the third part, after large pro recovery disk.
"Kung Fu in poetry", the same, large database activity and promote stable, smooth operation, the main work on the preparation before the big promotion, so preparation is the key.
A. Preparatory work before the big promotion
1. big promotion activities should get to know as much as possible, to become familiar with. Including business models, business processes, and a large volume of orders that may arise to promote, estimated peak, the estimated peak time, if there is explosion models commodities. In addition, should also participate in this great event to promote an understanding of the participants, especially the major participants of the IT department to ensure inter-departmental co-precise and smooth.
2. combing big promotion activities used in system link, there is a clearer understanding of the systems and applications on a link, database link flow chart full of big promotion event production.
3. carding database resources on the link. Perfect as a sort Excel, including the name of the database, data types, database version, uses, mainly supported by the system, DBkey, physical IP, virtual IP and other databases Size, disk space and available space, memory, maximum connection limit.
4. Check link database resiliency. Mainly full backup, log backup Job check availability check and backup files.
5. Check the link check the availability of the database. Is determined primarily highly available database architecture employed, the number of nodes, the node configuration, availability monitoring, status monitoring, monitoring synchronization.
6. Learn from the use of database nodes, usually the main attention during the forecast and the big promotion from latency issues, and the impact they may cause delays; whether optimization; and when the long delay occurs during big promotion, the availability of alternatives (For example, can a virtual IP on the node to drift from the primary node).
7. During the Great custom database to promote large-screen monitor, the main achievement of the basic realization of monitoring of all the key indicators database through a link on the full monitoring interface. (Monitoring the company's database is mainly achieved by Zabbix)
8. The pressure measurement for the link. Pressure measurement process should pay special attention to the following indicators: TPS, transaction response time, number of successful transactions, CPU, memory and disk usage and other servers. For the case of databases, pressure measuring points can identify bottlenecks and optimize more targeted. Further, there is a pressure test is to assess the function of the maximum performance of the system. For maximum performance, do the front end of a traffic restrictions, especially in the merchandise display, shopping cart, payment and other functions. Traffic restrictions, both to ensure the user experience, but also to prevent past data requests Cache, DB to drag down.
9. (For example: Zabbix) was observed by monitoring tools for each database server resource consumption. Recommendations observe the operation of the last week, such as CPU, memory fluctuations, the peak number of connections, is reasonable.
10. Adoption of the monitoring tools, such as the slow query log of SQL statements consuming more resources to sort out targeted optimization. Conventional optimization methods are: creating a new index, the index adjustment, data archiving, with or without a large field, update the table structure, data archiving, SQL statement optimization.
11. The link data delay monitoring. The main reason for the delay may be too long or the request queue receiving network delay impact, paying particular attention to this time span room (cross IDC) and the application requests data synchronization.
Application deployment likely impact of changes to the database caused during the 12 big promotion assessment. For example, in order to promote the activities of the system to deal with large requests, SA may increase the deployment of applications.
13. During the high-actuation database performance threshold estimate. Reasonable threshold is an accurate measure of the health of the database under the circumstances thermometer big promotion.
14. The application may sort degraded. For example, the data archived Job pause, BI and other data extracted Task delay.
15. prepare contingency plans. Contingency plans should be as detailed as possible, so that my heart there is spectrum, hands feet. Plan should include: updating the needs of business data which is common DBA needs to participate in the spare physical resources which, for the operation command might be used to repair the fault, change and approval process exception handling, drift virtual IP operation command. Spare physical resources required to refine the list to the server type, operating system, resources, specifications, pre-installed systems, IP and so on.
16.DBA duty planning.
II. Promote big row
1. Note that the database system monitor real-time monitoring.
2. The link data latency monitoring.
3. Inspection of the main database and server nodes.
4. promote development to keep abreast of the big cases, especially orders.
The demand for change should be particularly cautious.
6. Record the main anomalies that occur during big promotion.
III. After the big promotion Replay
Figure 1. perfect complement link to promote the use of large, complete node did not expect.
2. collect aggregate problems arise during the big promotion.
3. one by one on the issues that arise during large pro recovery disk, to find solutions to optimize and keep track of.
4. After the big promotion activities, the need for timely restore degraded service.
This article belongs to the author, without the author's consent shall not be reproduced , thanks! ! !
This article belongs to the author, without the author's consent shall not be reproduced , thanks! ! !
This article belongs to the author, without the author's consent shall not be reproduced , thanks! ! !