Hello, Is this allready fixed on svn version? With old bird we had problem that some times it leave ospf serrions to init state. With 1.3.5 bird looks that there is still problem that if remote ospf peer get hanged "right way" (on this case guagga node) bird leave session to init state. Naturally this is quagga problem and routes behid that node is dead anyway. BUT looks that if any of ospf sessios are down it can disturb routes from other sessions on same interface. Other sessions on same interface go to 2way/other state and routes are lost also behind of them. So if we have this kind configuration rA ----- rB | SW / \ netC-rC rD - netB | netA And if session to rC hangs to init state it cause that rD go to 2way/other and also routes to netB and netA are lost. On normal state: Router ID Pri State DTime Interface Router IP 10.123.123.99 1 full/other 00:01 eth3 y.y.y.1 10.123.123.211 1 full/other 00:10 eth3 y.y.y.212 10.123.123.1 1 full/other 00:10 eth3 y.y.y.210 y.y.y.102 1 full/dr 00:10 eth3 y.y.y.102 x.x.x.253 1 full/other 00:10 eth3 y.y.y.17 c.c.c.5 1 full/dr 00:09 br70 z.z.z.226 On problem state: Router ID Pri State DTime Interface Router IP 10.123.123.211 1 2way/other 00:10 eth3 y.y.y.212 10.123.123.1 1 init/other 00:09 eth3 y.y.y.210 x.x.x.253 1 full/bdr 00:10 eth3 y.y.y.17 y.y.y.102 1 full/dr 00:10 eth3 y.y.y.102 10.123.123.99 1 2way/other 00:10 eth3 y.y.y.1 c.c.c.5 1 full/bdr 00:10 br70 z.z.z.226 -- F-Solutions Oy Tapio Haapala PL7, 90571 Oulu GSM 0400 998371 Skype burner- IRC Burner@ircnet
On Mon, Apr 02, 2012 at 09:40:19PM +0300, Tapio Haapala wrote:
Hello, Is this allready fixed on svn version?
With old bird we had problem that some times it leave ospf serrions to init state. With 1.3.5 bird looks that there is still problem that if remote ospf peer get hanged "right way" (on this case guagga node) bird leave session to init state. Naturally this is quagga problem and routes behid that node is dead anyway. BUT looks that if any of ospf sessios are down it can disturb routes from other sessions on same interface. Other sessions on same interface go to 2way/other state and routes are lost also behind of them.
So if we have this kind configuration
rA ----- rB | SW / \ netC-rC rD - netB | netA
And if session to rC hangs to init state it cause that rD go to 2way/other and also routes to netB and netA are lost.
If the rC disappears and it was designated router (full/dr in state field, it may be a different neighbor in each 'run') before, the new DR have to be elected. Did the condition solved itself (rD become full/*) after a while? -- Elen sila lumenn' omentielvo Ondrej 'SanTiago' Zajicek (email: santiago@crfreenet.org) OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net) "To err is human -- to blame it on a computer is even more so."
7.4.2012 15:46, Ondrej Zajicek kirjoitti:
On Mon, Apr 02, 2012 at 09:40:19PM +0300, Tapio Haapala wrote:
Hello, Is this allready fixed on svn version?
With old bird we had problem that some times it leave ospf serrions to init state. With 1.3.5 bird looks that there is still problem that if remote ospf peer get hanged "right way" (on this case guagga node) bird leave session to init state. Naturally this is quagga problem and routes behid that node is dead anyway. BUT looks that if any of ospf sessios are down it can disturb routes from other sessions on same interface. Other sessions on same interface go to 2way/other state and routes are lost also behind of them.
So if we have this kind configuration
rA ----- rB | SW / \ netC-rC rD - netB | netA
And if session to rC hangs to init state it cause that rD go to 2way/other and also routes to netB and netA are lost. If the rC disappears and it was designated router (full/dr in state field, it may be a different neighbor in each 'run') before, the new DR have to be elected. Did the condition solved itself (rD become full/*) after a while?
Hmm I see some problem on my picture... rC and rD is ment to be on same L2 with SW and rB. Problem is that rC does not disappear at all. It stays on init state. This condition continue if you do not manually kill that rC. Looks that there is no any kind timeout and when one router is on init state it does not start election on that L2. So when one router buggy software fails, another routers leave state where that failed router is on init/other state and another routers are on 2way/other state. I see that this actual problem is on that failed router but on critical enviroment I think ospf must be failproof and on this kind situation I think election need start without waiting that one failed router. I dont know how it is possible without breaking ospf protocol specs, but I think someone figure way ;) -- F-Solutions Oy Tapio Haapala PL7, 90571 Oulu GSM 0400 998371 Skype burner- IRC Burner@ircnet
participants (2)
-
Ondrej Zajicek -
Tapio Haapala