Hello all.

I don't know if anyone has this problem that is very wierd.

In the company where I work we have a ServerU L-800 with 16GB RAM running Quagga 0.99.22.4 (official package distro) and Keepalived acting like your main border router. We have three BGP peering receiving full route advertisement.
Three weeks ago, more or less, we passed through a problem that ksoftirq using 100% of one CPU core and even we have another seven free cores we lose 60% of your bandwith and ours custumers args that they access was very slow.
We have another exactly machine server as secundary router, so we reboob the primary, the secundary runned fine and we back to first. Apparently everything works fine. Pass one week we had tha same problem but this time we just stop the Keepalived service, the secundary assumed and after 40 minutes the master ksoftirq back to normal.
We let this way and antil know, this pass eigth days, is working well.
Looking for this on the Internet I just found out old anwers about bugs or arp storm. But with we have two iguals machines and just one have this problem I don't believe that is bug.
Anyone passed for this problem?

I have a RHEL 7, kernel 3.10.0-123.el7.x86_64, running Quagga as my border router.

Some times the ksoftrqd take about 100% of one processor, a heve other six and my kernel have the SMP enabled, but my custumers fill slow connection.

I looking for this kind of problem and coudent find any answer. Any one have pass this situation?

Here my hardware: ServerU L-800 Processor: Intel® C2758 "Rangeley" 8x2.41Ghz (Octa Core) Embedded c/ suporte AES-NI Memmory: 16 GB RAM Disk: 60GB SSD NIC: 6x Intel Gigabit Teams: Tow teams eatch with three interfaces.