ECS performance exception checks （Serialized）
Created#More Posted time:Dec 8, 2016 10:01 AM
If you encounter CPU full load during the usage of ECS, we recommend the following checks:
a) Check whether the maximum thread count of the program is sufficient;
b) Is the program code insufficiently optimized? Does it have endless loops or deadlocks?
c) Are the web configuration file parameters well optimized?
d) View the web and system logs to check for abnormal access;
e) Is the website leeched?
f) Are there currently any search engine crawlers crawling large areas of your website?
g) Is the website under small-scale cyber attacks? Are there abnormal processes?
h) Check your machine for viruses or Trojans;
For a Linux server, you can use the system and web logs, as well as the top, free, uptime, sar, and ps commands to identify the cause. For a Windows server, you can analyze the situation using the resource monitor.
[Elizabeth edited the post at Dec 9, 2016 9:18 AM]
1st Reply#Posted time:Dec 8, 2016 10:48 AM
If a Linux OS server halts or gets stuck, you can initiate overall troubleshooting on the following items:
a. Check whether the server processes and services occupy too much memory or fail to release the memory. After its memory is used up, the server will go down very quickly;
b. Check your /var/spool/cron for any cron jobs (scheduled jobs, automatic jobs) being executed in the corresponding periods;
c. Check whether the parameters of your web servers exceed the server performance ceilings, such as a too high value for maximum connections allowed;
d. Check whether there are too many processes, services are paralyzed or the server is suspended (for example, there are too many zombie processes that keep occupying the resources);
e. Run the cat /var/log/message command to check for any exceptions in the system log;
f. Check for any bad blocks on the disk; (You can use the badblocks command in combination with the parameter.)
g. The kernel consumption is too high. Check whether there is instantaneous excessive occupation of resources by a process or service;
h. Run the ps -ef command to check for potential process exceptions and attacks.
2nd Reply#Posted time:Dec 9, 2016 9:16 AM
Notes on server monitoring performance:
1) When the user-mode CPU usage is high, it means applications on the server need a large amount of CPU overhead, such as a number of queries and sorting computations of the database server.
2) When the kernel-mode CPU usage is high, it means the server spends a lot of time in scheduling or system calling.
3) If the IOWait utilization is high, it means most of the time the CPU is used waiting for disk I/O operations. At this time, you should indeed check whether your disk I/O is too high.
4) When memory is insufficient, the server will get slower and slower during usage, and you need to upgrade your memory.
3rd Reply#Posted time:Dec 9, 2016 13:19 PM
The server performance is okay. Why is the website loading slow?
Slow website loading is not a problem of the page. You should troubleshoot the code and server:
a) Request a static page and check the speed.
b) Check the SQL statements reading the database for any potential necessary optimizations.
c) Test the sqlconnection speed.
d) Check the code for any endless loops.
e) Check whether there are code blocks for downloading data from other websites. If yes, check whether the slow loading of other websites is the cause.
4Floor#Posted time:Dec 12, 2016 11:21 AM
What should I do if the 500 error is reported for the website on the server?
We suggest you:
a) Check whether the personal center pages are fully uploaded.
b) Check whether the database is missing some lists.
c) Check whether the space is normal, and whether the personal center page path is correct.
d) Check the read/write permissions of the website file.
e) Check whether there are errors in the directory-changing syntax, such as redirection.
f) Check whether there is a program error.
g) Check whether the website is infected with viruses or Trojan horses.
5Floor#Posted time:Dec 12, 2016 14:54 PM
What should I do if the 502 error is reported for the website on the server?
We suggest you check the following:
a) Whether there are exceptions in concurrent connections and processes on your server.
b) Whether there are issues with your IIS or php-cgi, php-fpm process configurations, and whether the parameter settings should be optimized.
c) Whether slow queries to the database exist in your website program or database.
d) Whether there are exceptional situations in your web server and database server logs.
e) The server performance and bandwidth utilization.
6Floor#Posted time:Dec 13, 2016 10:22 AM
How can I solve the packet loss in testing with ping command?
We suggest you check the following:
a) The local network.
b) Whether the server bandwidth is normal and whether the load is sometimes too high;
c) Whether there are issues with firewall settings of the user server;
d) Whether the website is infected with viruses (we recommend you upgrade all patches promptly, detach unnecessary services and disable unnecessary ports);
e) Whether there is a delay in the intermediate network, by tracking the route.
f) Whether the server is under attack.
7Floor#Posted time:Dec 14, 2016 11:41 AM
Users reflect the delay variation with the ping command to the ECS changes a lot - it may be 20 ms at this moment and 200 ms at the next. What’s the cause?
(1) First check the server’s performance. If the CPU is at full load, the problem might be caused by any of the following factors: there is a high load; the bandwidth is at full capacity; the memory usage is high; there is a virus; or a process does not run properly.
(2) Ping other local websites to see if they have the same issue. This can tell you if the local network is unstable.
(3) If the ping command can be successfully sent to other websites, the ping or route tracing graphs from the local server to the ECS during the instable period can be provided by the user to help identify whether the problem is caused by any intermediate routing node.
(4) Submit a ticket to check whether there is a network fault or network instability in the server room.
8Floor#Posted time:Dec 15, 2016 9:55 AM
Users reflect the remote connection to the Linux ECS is normal, but login to the ECS is slow in response and the server is not fluent. Why?
We suggest you:
1) The first task is to identify whether the problem originates from the server or the network. First, check whether the user monitoring performance is normal. You can run the uptime command to show whether the average load of the server is too high, run the free command to see whether the memory usage is normal, and run the iostat command to see whether the disk IO is a bottleneck, and run the top command to check whether there are exceptions in the processes.
2) If the user host performance is normal, the user should provide the trace route tracing graph and ping value graph from the local server to the ECS. Check from the node checks and delays in the routing graph whether the problem is from the intermediate routing nodes or the user’s local network.
9Floor#Posted time:Dec 16, 2016 9:53 AM
What if the server bandwidth is full?
A high traffic volume may be caused by the following factors:
a) Large websites (such as portal sites and online shopping sites): These websites handle heavy access traffic and generate a high page view, a high hits value, and a large daily traffic volume.
b) Poor webpage design: If webpages contain large images or audio/video files, this can make the webpages too big.
c) Webpage content leeching or data downloads: The websites provide *.mp3, *.rar, *.zip, *.exe files for download, or stream video or audio files.
d) Websites under DDoS attacks or infected with Trojans: This can produce high traffic volumes in a short time.
10Floor#Posted time:Dec 20, 2016 14:03 PM
How to avoid high traffic volumes:
a) If your site is relatively large, and website click rates are high - You can upgrade your bandwidth.
b) We suggest you reduce the number of audio/video files.
c) Reasonably arrange the webpage layout.
d) Enhance website protection.