Valérie P
2016-05-13 15:51:42 UTC
Hi,
I am testing the Strongswan solution between two servers in my lab.
OS : Ubuntu 14.04
StrongSwan : 5.1.2
I observed today weird behaviour of our StrongSwan server when an even IP
address is used on the WAN interface. Half of my CPUs are full when I used
10.0.0.2 on my WAN interface, and it falls back to 0.3% of use when the IP
address is 10.0.0.1 (odd). Same results with 10.0.0.3.
I tried different mix (even to odd, odd to odd, even to even) on both
servers. The behaviour is reproducible.
Random information :
- irqbalance is enabled
- servers are AESNI capable
Did any of you ever experience such behaviour? Should the use of an odd IP
address on the WAN interface be considered as a prerequisite?
Regards,
Val
I am testing the Strongswan solution between two servers in my lab.
OS : Ubuntu 14.04
StrongSwan : 5.1.2
I observed today weird behaviour of our StrongSwan server when an even IP
address is used on the WAN interface. Half of my CPUs are full when I used
10.0.0.2 on my WAN interface, and it falls back to 0.3% of use when the IP
address is 10.0.0.1 (odd). Same results with 10.0.0.3.
I tried different mix (even to odd, odd to odd, even to even) on both
servers. The behaviour is reproducible.
Random information :
- irqbalance is enabled
- servers are AESNI capable
Did any of you ever experience such behaviour? Should the use of an odd IP
address on the WAN interface be considered as a prerequisite?
Regards,
Val