Project

General

Profile

Issue #2328

softIRQ maxed-out to 100% only on one core with ipsec

Added by Yogesh Mantri over 8 years ago. Updated over 7 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Noel Kuntze
Category:
configuration
Affected version:
5.5.1
Resolution:
No feedback

Description

Hi Strongswan team,

PFA ipse.conf, RPS configuration and pcrypt configuration I used.

I have total 8 cores CPUs and 8192MiB RAM allocated to Centos VM.
I have enabled RPS on least 4 cores and dedicating pcrypt most 4 cores.
Pcrypt using it four cores correctly. But only once core is being used from RPS.
Also I see, softIRQ not being distributed among core 0,1,2 and 3.

Could you please assist to understand any way to distribute ipsec NIC/softIRQ traffic among multiple cores when ipsec is enabled?

strongswan_ticket_info.tgz (4.4 KB) strongswan_ticket_info.tgz Yogesh Mantri, 18.05.2017 10:27

Related issues

Is duplicate of Issue #3122: Strongswan software iteruptsFeedback

History

#1 Updated by Noel Kuntze over 8 years ago

  • Status changed from New to Feedback
  • Priority changed from High to Normal

Unless you have set up RSS with multiple RX and TX queues, only one core will receive softIRQs. But even then, the processing of the ESP packets of one SA will not be parallelized. If you want to increase the throughput between two sites, configure several CHILD_SAs, one for each core, bind the SAs to the queues and use marks on both sides to tell them apart. Then divide traffic among the SAs using marks.

#2 Updated by Noel Kuntze over 8 years ago

  • Assignee deleted (Tobias Brunner)

#3 Updated by Tobias Brunner over 7 years ago

  • Status changed from Feedback to Closed
  • Assignee set to Noel Kuntze
  • Resolution set to No feedback

#4 Updated by Tobias Brunner about 6 years ago

  • Is duplicate of Issue #3122: Strongswan software iterupts added