Hi Jie, Inline.
> On Dec 27, 2021, at 7:27 PM, Jie Deng <deng...@webray.com.cn> wrote: > > Hi Florin, > > Sorry, I don't describe the CPU model comprehensively. The CPU model I use is > Intel(R) Xeon(R) CPU E5-2670 v3 @ 2.30GHz, There are two such CPUs in my > device, So it's 24 cores. Ah, okay. Nonetheless, it’s a Haswell. > can CPS go even higher, in spite of the CLI issues? > Yes,however, when the CPS is increased to more than 50K, there will be > another problem, that is, there is not enough local port, As shown in the > following log: > Dec 28 10:54:52 ubuntu vnet[9489]: message repeated 71784 times: [ > session_mq_connect_one:195: connect returned: no lcl port] Yup, that’s expected if all the active opens are done to only one IP. If you’re running the server on the same host as the proxy, maybe try to use cut-through sessions (attach the server to the same vpp). > > We are adding virtual IP through cli TCP SRC address to solve this problem, > but there are also some problems with this cli. Anything beyond Mercury’s last patch? > > could you try changing max_connects in session_mq_handle_connects_rpc to > something like 64? does it change anything? > This can't change anything. That probably means the issue is related to the fact that the connect logic spends more time trying to find a free port. So the best solution is to just rely on the cli above to avoid port collisions. Regards, Florin > > Regards, > Jie > >
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#20674): https://lists.fd.io/g/vpp-dev/message/20674 Mute This Topic: https://lists.fd.io/mt/87933432/21656 Mute #vpp-hoststack:https://lists.fd.io/g/vpp-dev/mutehashtag/vpp-hoststack Mute #hoststack:https://lists.fd.io/g/vpp-dev/mutehashtag/hoststack Group Owner: vpp-dev+ow...@lists.fd.io Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-