What Can I Do If There Is Packet Loss?
Scenarios
This FAQ provides guidance for you to troubleshoot packet loss.
Troubleshooting
Check Item |
What to Do |
---|---|
Check whether client requests are blocked by access control policies. |
|
The Maximum Number of Concurrent Connections That IP as Backend Servers Can Handle |
Check whether the maximum number of concurrent connections that IP as backend servers can handle is less than the number defined by the load balancer specifications. |
Check whether backend servers are available by running the curl command instead of the ping command. |
Access Control Blacklist or Whitelist
- Symptom: Some specific IP addresses are blocked.
- Analysis: The IP addresses may be blocked by a blacklist or whitelist.
- A blacklist: Packets from the IP addresses in the blacklisted IP address group will be denied to access the load balancer.
- A whitelist: Packets from the IP addresses not in the whitelisted IP address group will be denied to access the load balancer.
- Solution: You can view the monitoring metrics Blocked Packets and Blocked Traffic to check whether there are packets blocked by blacklist or whitelist. For details about the metrics, see Monitoring ELB Resources. If there are blocked packets, check whether the clients are blocked by the blacklist or whitelist.
The Maximum Number of Concurrent Connections That IP as Backend Servers Can Handle
- Symptom: When ELB is used to route traffic across IP as backend servers, the number of concurrent connections does not reach the upper limit, but there are still access failures or dropped packets.
- Analysis: An IP as backend server can handle no more than 100,000 concurrent connections. If there is only a small number of IP as backend servers, the maximum number of concurrent connections that these servers can handle may be less than the number defined by the load balancer specifications.
- Solution: You can use a formula to calculate the maximum number of concurrent connections that can be handled by the IP as backend servers associated with a load balancer.
Assume that your load balancer is running in two AZs and has ten IP as backend servers. You can use either of the following formulas to calculate the maximum number of concurrent connections:
- A public network load balancer: 100,000 concurrent connections × Number of IP as backend servers × Number of AZs
In this example, the maximum number of concurrent connections is calculated as: 100,000 × 10 × 2 = 2,000,000.
- A private network load balancer that is in the same AZ as the clients:
100,000 concurrent connections × Number of IP as backend servers
In this example, the maximum number of concurrent connections is calculated as: 100,000 × 10= 1,000,000.
If the maximum number of concurrent connections allowed is less than the number defined by the load balancer specifications, you can add more IP as backend servers.
- A public network load balancer: 100,000 concurrent connections × Number of IP as backend servers × Number of AZs
Backend Server Connectivity
- Symptom: The IP address of a load balancer can be pinged, but there are still access failures or dropped packets.
- Analysis: If the load balancer is working normally, it returns ICMP Echo to the ping command. However, ping packets are not forwarded to backend servers, so the ping command output does not reflect the actual status of backend servers.
- Solution: Check whether backend servers are available by running the curl command.
Feedback
Was this page helpful?
Provide feedbackThank you very much for your feedback. We will continue working to improve the documentation.See the reply and handling status in My Cloud VOC.
For any further questions, feel free to contact us through the chatbot.
Chatbot