Recently, when I used a browser to access this site, I tried to refresh it a few more times, and it would get a 503 error, as shown in the figure below:
503 Service Temporarily Unavailable Meanwhile, the response header is as follows:
content-length: 608
content-type: text/html
date: Sat, 07 Jan 2023 13:50:32 GMT
via: HTTP/2.0 SLB.205 503 Service Temporarily Unavailable error
503 is an HTTP status code due to temporaryServer maintenance oroverload, the server is currently unable to process the request. This condition is temporary and will be restored after some time.
Generally speaking, the 503 error is mostly due toHigh website traffic, resulting in an error caused by excessive traffic or a large number of concurrent resources.
via response header
We see it in the 503 responsevia: HTTP/2.0 SLB.205The response header information, regarding the via response header, is explained as follows:
Via
Lists which proxy servers the response from the client to OCS or vice versa, and what protocol (and version) they sent the request with.
When the next proxy server receives the request from the first proxy server, it will copy the Via header of the previous proxy server's request in its own request, and add its relevant information to the back, and so on, when OCS receives the request from the last proxy server, it checks the Via header to know the route through which the request passes. For example: Via:1.0 236-81.D07071953.sina.com.cn:80 (squid/2.6.STABLE13) Based on the description, we can roughly guess that it isThe SLB (load balancing) service returned a 503 error directlyThat is, the request did not reach the real server on our backend, and the request was directly responded to by Alibaba Cloud SLB (Load Balancing) to the user.
At this time, I suddenly thought that Alibaba Cloud's SLB (load balancing) service has different specifications, and different specifications also have different concurrency limits, taking the simple type I (slb.s1.small) as an example.The maximum number of connections supported by this specification is 5000, new connections (CPS): 3000, and queries per second (QPS): 1000。 As shown below:
Solution:Upgrading SLB configurations!! Upgrading SLB configurations!! Upgrading SLB configurations!!
Regarding looking at SLB (load balancing) monitoring, it can be seen that it has exceeded the specification limit.
SLB configuration SSL will enable http/2.0 by default, because http2.0 will reuse TCP connections, and after a TCP connection is established, it will only be loaded on one node of SLB
The configuration of the performance-guaranteed slb.s1.small is as follows: Number of connections: 5000, CPS: 3000, QPS: 1000
The QPS of this specification is 1000, but the QPS of a single node of SLB is 1000/(8-1), the seventh layer is 8 nodes, and the QPS of a single node is around 142. http/2.0 is placed on the backend to configure
You can refer to ithttps://help.aliyun.com/knowledge_detail/55193.htmlexplanation of why each connection does not reach bandwidth peaks; The two are the same reason Resources:
The hyperlink login is visible.
The hyperlink login is visible.
|