Baidu academic revision and CPU 100% related?

After the release of the new version of the website, specially click the commonly used services, normal operation, everything restored as before.

However, not long after (less than 10 minutes), the whole system appeared no response again. After logging into the server to check, IT was found that the CPU was 100%, which was very strange. At that time, it was really suspected that the program had a vulnerability similar to Java deserialization, which was implanted into the mining program. But the odds were low, so I decided to dig a little deeper.

Search on Baidu: Windows CHECK CPU usage, and see an article: How to find Java applications in Windows that occupy too much CPU, very fit the current predicament.

Check whether the CPU usage of Java applications in Windows is too high

The above article has been described in more detail, but there is a part of the more redundant, I am making a small summary.

The installation Process Explorer

From Microsoft’s official website technet.microsoft.com/en-us/sysin… Download the installation package. The installation package is a zip file and decompressed to a customized folder.

After decompressing, run procexp64.exe directly. The interface is as follows:

Find threads that occupy the CPU

As you can see from the above figure, the application has taken up about 50% of the CPU. Because the CPU is dual-core, one request is currently stuck, so it takes up about half of the CPU. Select the application, right click, and then select The Properties menu to open the application Properties screen. In the properties screen, you can see the resources occupied by all threads of the program.

Export stack information using jStack

Jstack is a stack tracing tool provided by the Java VM. In some cases (the bin directory of the JDK is not configured with the PATH environment variable), jStack cannot be directly used on the command line. The most direct way to use jStack is to go to the bin directory in the Java installation directory. Exe 2380 >c:/stack.log, 2380 is the pid of the website application, and the last is the path to the stack log.

Look up the problem code from the stack

Open stack log C :/stack.log and search for 0xf70 and 0x1398:

To solve the problem

See the problem code, sure enough, or regular matching problem, but this re is really no problem, do not find the regular expression problem for a long time.

Finally found the root cause of the problem: ** Baidu academic interface is sometimes the new version, sometimes the old version, ** two kinds of regular expression can be applied to two kinds of pages, but for the pages that are not correct, it takes a long time to match the results, it is really pit to the extreme.

Before finding the problem, the solution I adopted was to simplify the above long re. Simplification means that all elements were matched in one sentence before, and now we match them step by step: first match a single record, and then match the title, year, abstract, etc. In a single record; With step – by – step matching, there is no problem of excessive CPU usage.

conclusion

1. Don’t make regular expressions too long if you don’t have to, even if you’re an expert at regular expressions.
2. For matching of multiple records, a single record is matched first (to determine the boundaries of each record), and then detailed fields are matched in the text of each record;
3. You can use Html parsing tools and use XPATH or CSS selectors instead of regular expressions.

Any Code, Code Any!

Scan code to pay attention to “AnyCode”, programming road, together forward.