Hi Helio,
Thanks for getting in touch.
About the Nokia issue with MD5: while i have (in)direct reports of MD5
working with other major vendors, ie. Cisco, Juniper and Huawei, i can't
confirm Nokia. Willing to help, if you can confirm some additional
information here or by unicast email, like OS version and config.
With regards to the BGP flaps: are you using pmbgpd? That is, can you
confirm there is no NetFlow / IPFIX in the picture? Are the flaps
observed at startup or (randomly) at steady state? Can you also share
your daemon config? I am essentially trying to determine if you could
just spin a few pmbgpd in parallel (same / similar config) that would
load share the BGP sessions and make use of multiple CPU cores and
leverage built-in support for SO_REUSEPORT.
Paolo
On 5/6/24 18:39, Helio Barroso (hbarroso) wrote:
All,
My customer setup demands BGP sessions against ~280 Nokia and Cisco
routers.
We are facing two different issues:
1. Regarding Nokia sessions we haven’t been able to establish sessions
using TCP MD5 authentication.
Same authentication works fine against Cisco routers.
Do you know if there is any specific configuration to
support Nokia BGP sessions or I can take for granted that is a Nokia
configuration issue ?
Nokia BGP configuration is the one recommended by the vendor.
2. For both, Cisco and Nokia, we are facing flaps on BGP sessions. We
have configured so far 90 sessions.
Server CPU peaks have been observed. PMBGPD cpu consumption peaks are
seen from 20 to 43% of total CPU utilization.
We are considering decreasing the keepalives as an alternative to
decrease CPU, but 2 questions arise:
1. Where can I change BGP keepalive and hold timer on PMACCT side ?
2. Is there any CPU vs # PMACCT sessions baseline that I should consider ?
Regards
_______________________________________________
pmacct-discussion mailing list
http://www.pmacct.net/#mailinglists
_______________________________________________
pmacct-discussion mailing list
http://www.pmacct.net/#mailinglists
