Thanks, Doug (and the others, who replied earlier today). Indeed 50ms appears too low on the MX80. When I fail a link across which LSPs (primaries or backups) pass, even those events being handled by the RE can occasionally cause BFD to miss 3 hellos, shutting down the OSPF neighbour. I've tried at 100ms with better result, though your recommendation of 150ms is well received. Do we know if this is something that *will* eventually be distributed to the PFE in future releases on the MX80?
David On 3 March 2011 15:27, Doug Hanks <dha...@juniper.net> wrote: > We generally recommend 150ms to most customers. The added benefit of going > from 150ms to 50ms is generally not enough to warrant the move. > > -----Original Message----- > From: juniper-nsp-boun...@puck.nether.net > [mailto:juniper-nsp-boun...@puck.nether.net] On Behalf Of Andy Harding > Sent: Thursday, March 03, 2011 10:07 AM > To: David Ball > Cc: juniper-nsp@puck.nether.net > Subject: Re: [j-nsp] BFD timers for OSPF - MX80 - 10.3R2.11 > > We are using bfd on mx80 with 300ms timers and no problems. Only 2 or 3 > sessions per box however. > > -- > > Regards > > Andy Harding > Internet Connections Ltd > > Phone: 0870 803 1868 > Mobile: 07813 975459 > Fax: 0870 803 1781 > Web: www.inetc.co.uk > Email: a...@inetc.co.uk > > On 3 Mar 2011, at 17:53, David Ball <davidtb...@gmail.com> wrote: > >> Ah, that might help explain it. And shame on me for not checking >> 'sh pfe statistics traffic protocol bfd', which of course shows none >> received or absorbed. >> I'll only have 2 sessions on each MX80, so I think I might leave it >> enabled, but may toy with the interval. I'm expecting the control >> plane to be kinda bored on these guys, so we'll see what it can >> handle. >> Thanks, Egor. >> >> David >> >> >> On 3 March 2011 10:42, Egor Zimin <les...@gmail.com> wrote: >>> Hello, David >>> >>> It looks like BFD implementation in MX80 is not distributed. At this >>> moment I have a case in JTAC. The case is opened yet, however, it >>> _looks_like_ bfd is not distributed. >>> Probably because of this BFD echomode is not supported. And using 30ms >>> timers for BFD ControlPackets can be not so easy task for RE's CPU. >>> >>> Because of this I don't see much sense to use BFD on MX80 at this moment. >>> >>> 2011/3/3 David Ball <davidtb...@gmail.com>: >>>> MX80s running 10.3R2.11 >>>> >>>> For those of you using BFD for OSPF, how low have you been able to >>>> set your minimum-interval timer? I have a pair of MX80s connected via >>>> XFPs and 1m patch cables and with my hellos set to 30ms and multiplier >>>> set to 3, I'm seeing failures. I haven't disabled distributed ppm. >>>> Moving to 50ms hellos seems to settle things down. The reason I'm >>>> wondering why I can't get away with lower timers is because when >>>> Juniper proof-of-concepted (yeah, that's a verb) Trio for us (albeit >>>> using MX960s), they used 15ms hellos with a multiplier of 3. >>>> >>>> Mar 3 10:06:06 router bfdd[1129]: BFDD_TRAP_STATE_DOWN: local >>>> discriminator: 1, new state: down >>>> Mar 3 10:06:06 router rpd[1257]: RPD_OSPF_NBRDOWN: OSPF neighbor >>>> 172.16.1.22 (realm ospf-v2 xe-0/0/2.0 area 0.0.0.0) state changed from >>>> Full to Down due to InActiveTimer (event reason: BFD session timed out >>>> and neighbor was declared dead) >>>> >>>> >>>> me@router> show configuration groups bfd-defaults-core-ospf >>>> protocols { >>>> ospf { >>>> area 0.0.0.0 { >>>> interface <*> { >>>> bfd-liveness-detection { >>>> version automatic; >>>> minimum-interval 30; >>>> multiplier 3; >>>> full-neighbors-only; >>>> } >>>> } >>>> } >>>> } >>>> } >>>> >>>> me@router> show configuration protocols ospf area 0.0.0.0 >>>> interface lo0.0 { >>>> passive; >>>> } >>>> interface xe-0/0/2.0 { >>>> apply-groups bfd-defaults-core-ospf; >>>> node-link-protection; >>>> } >>>> interface xe-0/0/3.0 { >>>> apply-groups bfd-defaults-core-ospf; >>>> node-link-protection; >>>> } >>>> >>>> >>>> David >>>> _______________________________________________ >>>> juniper-nsp mailing list juniper-nsp@puck.nether.net >>>> https://puck.nether.net/mailman/listinfo/juniper-nsp >>>> >>> >>> >>> >>> -- >>> Best regards, >>> Egor Zimin >>> >> _______________________________________________ >> juniper-nsp mailing list juniper-nsp@puck.nether.net >> https://puck.nether.net/mailman/listinfo/juniper-nsp > > _______________________________________________ > juniper-nsp mailing list juniper-nsp@puck.nether.net > https://puck.nether.net/mailman/listinfo/juniper-nsp > _______________________________________________ juniper-nsp mailing list juniper-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/juniper-nsp