One BGP session flapping but not the other
Hi, I have a pretty simple design: two edge routers, one hypevisor. The routing is done with the help of BGP. regis is the hypervisor. +---------+ +-------+ | nominoe | | budic | +---------+ +-------+ \ / \ / \ / AS204092 --------\-------/--------- \ / AS65000 +-------+ | regis | +-------+ The IPv4 session between nominoe and regis is flapping, but not the one with budic. The IPv6 sessions are stables. nominoe and budic are FreeBSD 11.1-RELEASE boxes, running bird 1.6.3 regis is a gentoo box running bird 1.6.3 too At each time the session is flapping, I can see 'bird: Kernel dropped some netlink messages, will resync on next scan.' in the logs of regis. What does that mean, and what part of the configuration is acting on it? The sessions are configured the same maner at the edge side: template bgp DOWNSTREAM { export all; import keep filtered; } protocol bgp bgp_alarig from DOWNSTREAM { local as 204092; description "BGP for alarig"; neighbor 89.234.186.100 as 65000; import keep filtered; import filter bgp_filter_customer_alarig_in; } And the same goes for the hypervisor: template bgp UPSTREAM { local as myasn; # Be able to see filtered routes with "sh route filtered" command import keep filtered; # Protect ourselves from massive routes leaks receive limit 700000 action block; # Announce only our IP allocations export where proto = "static_alarig_allocations"; export limit 1 action block; } protocol bgp bgp_nominoe from UPSTREAM { description "nominoe"; neighbor 89.234.186.7 as 204092; # Local address we use to establish the BGP session source address 89.234.186.100; # password='' # Accept all routes from Cogent except bogons and other bad stuff import filter bgp_filter_nominoe_in; } protocol bgp bgp_budic from UPSTREAM { description "nominoe"; neighbor 89.234.186.6 as 204092; # Local address we use to establish the BGP session source address 89.234.186.100; # password='' # Accept all routes from Cogent except bogons and other bad stuff import filter bgp_filter_budic_in; } I can paste the rest of the configuration if you think it’s relevant. Thanks, -- alarig
Hi, I don’t know if it helps, but here is the netlink options compiled in my kernel: regis linux # grep -i netlink /usr/src/linux/.config CONFIG_NETFILTER_NETLINK=y # CONFIG_NETFILTER_NETLINK_ACCT is not set # CONFIG_NETFILTER_NETLINK_QUEUE is not set CONFIG_NETFILTER_NETLINK_LOG=y CONFIG_NF_CT_NETLINK=y # CONFIG_NF_CT_NETLINK_TIMEOUT is not set # CONFIG_NETFILTER_NETLINK_GLUE_CT is not set # CONFIG_NETLINK_DIAG is not set # CONFIG_SCSI_NETLINK is not set -- alarig
Hi, By digging a bit more, I found that the issue seems to be with CARP. I found that the mastership is flapping between the routers as budic is a bit overloaded: Dec 22 23:42:32 nominoe kernel: carp: 4@em1.31: BACKUP -> MASTER (master timed out) Dec 22 23:42:32 nominoe kernel: carp: 4@em1.31: MASTER -> BACKUP (more frequent advertisement received) And, a each time the status change, the routes on the device protocol are reinitialised (the timestamp on 'show route' changes) For now, I forced nominoe to be the master and the session is stable again. Is it a known behaviour? -- alarig
participants (1)
-
Alarig Le Lay