On Mon, Jan 30, 2012 at 11:19:26AM +0100, Ondrej Zajicek wrote:
On Wed, Jan 25, 2012 at 03:14:10PM +0100, Bjoern Engels wrote:
we're running redundant pairs of routers using keepalived for VRRP and BIRD for OSPF. (bird-1.3.3 through bird-1.3.6 actually)
Static routes are configured in BIRD so when keepalived transitions to master state it adds ip addresses to a bunch of vlan interfaces and the routes' gateways become directly reachable and are being announced.
Now when I failover from one primary router to a backup router by stopping keepalived the interfaces' addresses are being deleted on the former master and added to the backup router.
So the failover would cause removal of these static routes?
Yes. When all addresses have been deleted on an interface it changes to DOWN so the routes' next hops become unreachable and the they are being withdrawn.
Immediately after such a transition all BIRD routers fill their log files periodically with this kind of messages:
Jan 25 14:47:06 cr4b bird: OSPF: LSA disappeared (Type: 0005, Id: 192.0.2.0, Rt: 10.10.10.170) Jan 25 14:47:06 cr4b bird: OSPF: LSA disappeared (Type: 0005, Id: 192.0.2.128, Rt: 10.10.10.170)
I guess these LSAs represents old static routes (Rt: 10.10.10.170 is the router that removed the static routes), is this true?
Correct.
Are there any problems with this other than log filling? You could check using 'show ospf state' and 'show ospf lsadb' whether the old LSAs/routes are properly removed and the new are properly installed during failover (and during the failover back to the primary).
Log filling (>> 100.000 entries per day*router) seems to be the only problem, "show ospf lsadb" shows LSAs that have been advertised by the active (master state) routers and LSAs of inactive/backup routers have been deleted.
proto/ospf/lsupd.c says this should never happen:
/* Probably flushed LSA, this should not happen */ log(L_WARN "OSPF: LSA disappeared (Type: %04x, Id: %R, Rt: %R)", lsr->lsh.type, lsr->lsh.id, lsr->lsh.rt);
Is this some bug in BIRD or caused by the way we use it?
This is probably some minor bug in LSA flushing, i will check that.
Alright, thank you! Regards Bjoern -- Bjoern Engels OpenIT GmbH engels@openit.de In der Steele 33a-41 PGP keyID 1024D/895F13C3 D-40599 Duesseldorf ________________________________________________________________________ HRB 38815 Amtsgericht Duesseldorf USt-Id DE 812951861 Geschaeftsfuehrer: Oliver Haakert, Maurice Kemmann