How to Troubleshoot DB FAQ

Under normal circumstances, the system will encounter some problems, then how do we locate the problem after the reason? Here I only say how to locate the db problem.

See this article has a premise: monitoring data to complete! monitoring data to complete!! monitoring data to complete!!!
Like the following.

At first glance, there is a performance jitter, how to know if the system is not a problem, can be informed by the following ways:

• Application log
• Monitoring alarms
• User perception

Whether it is monitoring the alarm, or user perception, in the final analysis, you have to return to the application, from the application log to find out which interface is the exception, the reason for the interface exception of the following several situations:

• System exceptions, such as exceeding the load
• Network problems, such as network card full, net packet loss
• IO problems, such as brush disk, IO scheduling algorithm setup problem
• File system problems, such as kernel bugs
• DB issues, such as those that contain the above, rotten SQL issues

If you say that the above several cases are not abnormal after troubleshooting, or through the log at a glance to find the DB problem, what is the problem? Normally, I would do this:

1. Open the Monitoring page to see "Performance"

If there is no monitoring, I have no way, only according to the application of the error guess.

Keep your eyes wide open:
1. View the overall performance of the problem point in time
2. Pass the exception to find the problem point
The truth of Medicine says: "Many breakthroughs in medicine begin with the study of exceptions." ”

Lao Luo also said, "in the ordinary study, the most precious things are not those in line with the theory of things, precisely those who look at strange, exceptional things, the existing theory is often explained, you must not ignore him, because it is the next innovation seedling." ”
Here is no exception, such as first look at the overall Qps,tps there is no change, the network has no jitter and so on, in short, is found and other time periods are not the same trend. This is the problem point, take the first picture for example, we see the RT high, in general, RT soaring High usually for the following reasons:

• Rotten SQL line above 90% basically that's the reason
• Hardware problems have appeared on the line, but not much
• File system Bug
• MySQL Bug
• Network problems

About rotten SQL In fact there are all kinds of, you can go to the Internet to search a lot of, here is not tired, after the time, I may write;

MySQL Bug if the elimination of rotten SQL, there are some weird problems can be suspected MySQL is a bug, you can view the bug list;

Again is the network problem, if the determination is not rotten SQL, the system of QPS is very high, if the system has occasional jitter, you can view network monitoring;

Finally, network problems and file system problems appear very few, but also do not rule out, only specific problems specific analysis;

How to Troubleshoot DB FAQ