One-time diagnosis of -- Application server bottleneck

Today, the on-site implementation feedback system is slow as a whole, and many complaints have been received from users asking the site to send back WebLogic logs and Oracle Database reports. Briefly describe the system architecture: the typical B/S three-tier architecture. The development language is Java, the middleware uses WebLogic, the database uses Oracle, and the PC server.

1. Analyze WebLogic logs and database reports. WebLogic logs contain no stuck requests, no connection pool problems, and no OS errors. There is no pressure on the Database Report.

2. the monitoring results of httpwatch must be retrieved from the site. It is found that three small requests in a large functional request are waiting for more than 799 S. One of the three small requests is to download an image, the other two theoretically cannot be slow, and cause later request timeout (error_internet_timeout ). Repeat the ideas:

Client browser <========> middleware (WebLogic) <========> database (Oracle)

There is no problem with the link from middleware to database. If there is a problem, it can be captured from the middleware log. There is a problem with the link from the client browser to the middleware. There are two possibilities: first, there is a problem with the network, and second, the application server does not respond.

3. check whether there are any problems with other systems. The answer is that there is no problem with other systems. The possibility of network problems is small. To know whether the application server has a problem, you can only monitor the OS metrics of the application server.

The operating system on the application server is exactly monitoring, how to configure, please refer to http://blog.csdn.net/stevendbaguo/article/details/8737743. It is nothing more than viewing the CPU, memory, physicaldisk three indicators, the monitoring data using EXCEL to generate images, and finally found that physicaldisk this indicator was very abnormal on the morning of the 21st, has been in the peak, the maximum value is 80 Mb/s. According to my simple storage knowledge, the 5400-to-54 Mb/s hard disk, the 7200-to-72 Mb/s hard disk, and the IBM small machine storage can reach 200 Mb/s, SSD can reach 400 Mb/s-500 Mb/s. The field is neither a small machine nor an SSD. I think 80 Mbit/s is very high and will produce a lot of Io waits.

4. After telling the analysis result to the implementation colleagues, the feedback is that the disk is backed up in the morning and a large number of files are copied. It seems that the cause has been found, and he is warned that this kind of thing will be done during off-duty time.

One-time diagnosis of -- Application server bottleneck