Here are the ab test results:
Before:

Server Software:        nginx/1.15.5
Server Hostname:        10.0.0.147
Server Port:            8080

Document Path:          /
Document Length:        612 bytes

Concurrency Level:      300
Time taken for tests:   120.017 seconds
Complete requests:      48983
Failed requests:        0
Total transferred:      41390635 bytes
HTML transferred:       29977596 bytes
Requests per second:    408.13 [#/sec] (mean)
Time per request:       735.051 [ms] (mean)
Time per request:       2.450 [ms] (mean, across all concurrent requests)
Transfer rate:          336.79 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0  722 457.9    999    3009
Processing:     0    9  23.3      1     424
Waiting:        0    9  23.2      1     424
Total:          0  732 449.6   1000    3201

Percentage of the requests served within a certain time (ms)
  50%   1000
  66%   1000
  75%   1000
  80%   1001
  90%   1002
  95%   1004
  98%   1007
  99%   1198
 100%   3201 (longest request)

After:

Server Software:        nginx/1.15.5
Server Hostname:        10.0.0.147
Server Port:            8080

Document Path:          /
Document Length:        612 bytes

Concurrency Level:      300
Time taken for tests:   106.312 seconds
Complete requests:      580000
Failed requests:        0
Total transferred:      490100000 bytes
HTML transferred:       354960000 bytes
Requests per second:    5455.62 [#/sec] (mean)
Time per request:       54.989 [ms] (mean)
Time per request:       0.183 [ms] (mean, across all concurrent requests)
Transfer rate:          4501.95 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0   28 392.7      0   63157
Processing:     5   25  88.6     23   25652
Waiting:        4   25  88.6     23   25652
Total:         15   54 409.7     24   63348

Percentage of the requests served within a certain time (ms)
  50%     24
  66%     25
  75%     26
  80%     27
  90%     30
  95%     38
  98%     56
  99%   1025
 100%  63348 (longest request)

/cc @m1093782566 @jsravn
/assign @m1093782566

@Lion-Wei: GitHub didn't allow me to request PR reviews from the following users: jsravn.

Note that only kubernetes members and repo collaborators can review this PR, and authors cannot review their own PRs.

Good improvement, THANKS!

/lgtm

/approve

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: Lion-Wei, m1093782566

The full list of commands accepted by this bot can be found here.

The pull request process is described here

/retest
This bot automatically retries jobs that failed/flaked on approved PRs (send feedback to fejta).

Review the full test history for this PR.

Silence the bot with an /lgtm cancel comment for consistent failures.

/milestone v1.13

/priority important-soon

Hello everyone：
We are very fortunate to tell you that this bug has been fixed by us and has been verified to work very well. The patch（ipvs: avoid drop first packet by reusing conntrack） is being submitted to the Linux kernel community. You can also apply this patch to your own kernel, and then only need to set net.ipv4.vs.conn_reuse_mode=1(default) and net.ipv4.vs.conn_reuse_old_conntrack=1(default). As the net.ipv4.vs.conn_reuse_old_conntrack sysctl switch is newly added. You can adapt the kube-proxy by judging whether there is net.ipv4.vs.conn_reuse_old_conntrack, if so, it means that the current kernel is the version that fixed this bug.
That Can solve the following problems:

1. Rolling update, IPVS keeps scheduling traffic to the destroyed Pod
2. Unbalanced IPVS traffic scheduling after scaled up or rolling update
3. fix IPVS low throughput issue fix IPVS low throughput issue #71114
   fix IPVS low throughput issue #71114
4. One second connection delay in masque
   https://marc.info/?t=151683118100004&r=1&w=2
5. IPVS low throughput IPVS low throughput #70747
   IPVS low throughput #70747
6. Apache Bench can fill up ipvs service proxy in seconds Support Restart policy in the kubelet (pre-design) #544
   Apache Bench can fill up ipvs service proxy in seconds cloudnativelabs/kube-router#544
7. Additional 1s latency in host -> service IP -> pod when upgrading from 1.15.3 -> 1.18.1 on RHEL 8.1 Additional 1s latency in host -> service IP -> pod when upgrading from 1.15.3 -> 1.18.1 on RHEL 8.1 #90854
   Additional 1s latency in host -> service IP -> pod when upgrading from 1.15.3 -> 1.18.1 on RHEL 8.1 #90854
8. kube-proxy ipvs conn_reuse_mode setting causes errors with high load from single client kube-proxy ipvs conn_reuse_mode setting causes errors with high load from single client #81775
   kube-proxy ipvs conn_reuse_mode setting causes errors with high load from single client #81775

Thank you.
By Yang Yuxi （TencentCloudContainerTeam）

Following-up on @yyx's comment above for posterity.

The patch mentioned above didn't make it to the kernel but there are two recently merged patches worth highlighting. One of them fixes the 1 second delay issue and the other fixes dropped packets when stale connection entries in the IPVS table are used:

1. http://patchwork.ozlabs.org/project/netfilter-devel/patch/[email protected]/
2. http://patchwork.ozlabs.org/project/netfilter-devel/patch/[email protected]/