Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

kni rx isolation using rte_flow #794

Merged
merged 5 commits into from
Jan 29, 2022
Merged

Conversation

ywc689
Copy link
Collaborator

@ywc689 ywc689 commented Jan 19, 2022

No description provided.

…vice bootup

Signed-off-by: ywc689 <ywc689@163.com>
The idea is to specify kni interface with an ip address, and isolate
all traffic target at the address to a dedicated nic rx-queue, which
may eliminate disturbances of dataplane when overload.

Note that not all nic can support this flow type under the premise of
sapool, and an additional rx queue is required, which is configured
into dpvs's kni worker.

Signed-off-by: ywc689 <ywc689@163.com>
@ytwang0320 ytwang0320 added the pr/codes-reviewed-ok code review passed and no problem found label Jan 21, 2022
@ywc689 ywc689 force-pushed the kni-rx-isolation branch 5 times, most recently from 2354cdc to 394c80b Compare January 21, 2022 10:40
@ywc689
Copy link
Collaborator Author

ywc689 commented Jan 21, 2022

This feature is derived from PR #695. Here are the major changes:

  • Adapt to DPVS v1.9, remove the obsoleted flow director apis.
  • Reduce hardware/driver dependencies, remove tricky and buggy logics for specified device types, and adjust rss with rss apis rather than rte_flow.
  • Add dpip tools for flow configuration.
  • Add supports for DPVS's bonding, vlan devices.

Signed-off-by: ywc689 <ywc689@163.com>
Signed-off-by: ywc689 <ywc689@163.com>
Signed-off-by: ywc689 <ywc689@163.com>
@ywc689 ywc689 added the pr/accepted the pr passed all review stages and await to be merged label Jan 29, 2022
@ywc689 ywc689 merged commit 8aad0cb into iqiyi:devel Jan 29, 2022
@githubljy
Copy link

@ywc689 您好 我测试使用最新devel分支包含 kni rx队列独立的这个merge ,通过dpip flow命令使管理口ip,OSPF流量走kni队列,经过实际压力测试发现在业务核流量达到瓶颈时候(压测单个lcore开始丢包,或多个lcore到达瓶颈都会影响)依然会影响到独立kni队列的管理口ip ,和OSPF流量导致时延、丢包(测试过程中通过dpdk-pdump抓包确认管理口ip,ospf只走kni队列)。我看merge给出的测试用例是通过在业务lcore上delay 1ms模拟得出验证结果kni rx独立后业务流量过载的情况下不影响kni流量,我也测试模拟过,模拟结果也一样不影响,但是目前在实际流量压力测试下还是会影响。这块目前有点怀疑是否是网卡rss相关底层影响(我测试使用的网卡是82599ES 10-Gigabit)?目前还没什么思路,所以请教一下,或者这块是否有经过实际流量压测的验证结果。

@ywc689 ywc689 deleted the kni-rx-isolation branch July 1, 2022 03:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
pr/accepted the pr passed all review stages and await to be merged pr/codes-reviewed-ok code review passed and no problem found
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants