lb of keepalive problem

0. keepalive

His intention is clear to everyone, that is, to reduce shaking hands three times, setting a timeout, for instance 20s, is not requested in the 20s, still maintained a connection, this time the request came, after a three-way handshake is not required to re-tcp, and if than will break, then we should re-connection 3-way handshake.

A normal reply keepalive head has two parameters:

  1. timeout
  2. max
HTTP/1.1 200 OK
Connection: Keep-Alive
Content-Encoding: gzip
Content-Type: text/html; charset=utf-8
Date: Thu, 11 Aug 2016 15:23:13 GMT
Keep-Alive: timeout=5, max=1000
Last-Modified: Mon, 25 Jul 2016 04:32:39 GMT
Server: Apache

(body)

7 1. Cloud vendors load balancing layer disposed

To give domestic manufacturers, for example,

  1. Ali cloud, the default keepalive_timeout 15s, can not be closed.
  2. Huawei cloud, the default keepalive_timeout 300s, can not be closed.

So it is clear that the value of Huawei's cloud a little too big.

2. keepalive timeout is too large due to the influence

That is, the client will stay long in the lb above, if you have only 1 lb, a computer room, in fact, nothing that impact, but if you have more room to deploy, users will not cause problems elsewhere.

Guess you like

Origin www.cnblogs.com/gqdw/p/11563086.html