I've observed this behavior on MX104 in both 13.3R4.6 and 14.1R5.5, without or without nonstop routing [rpd running on backup RE]. In general I avoid ISSU and we don't always force failover from RE1 to RE0 after maintenance, so RE1 isn't always our backup RE. For now I am using "syslog host $host log-prefix" under re0/re1.
If I one day pursue this further with Juniper, I'll report back my findings. -Michael > -----Original Message----- > From: juniper-nsp [mailto:juniper-nsp-boun...@puck.nether.net] On Behalf > Of Chris Hellberg > Sent: Wednesday, September 02, 2015 6:10 PM > To: juniper-nsp@puck.nether.net > Subject: Re: [j-nsp] backup RE and default-address-selection question > > Docs seem to indicate that the source IP of host-initiated traffic is > the fxp0 interface when you set default-address-selection and the egress > interface is fxp0. Lo0 is used if traffic goes out a routed interface > (e.g. when doing inline management). > > http://www.juniper.net/documentation/en_US/junos13.2/topics/reference > /configuration-statement/default-address-selection-edit-system.html > > If not, may be a bug; what's your code version? > > Another option that springs to mind and that's to put the [edit system > syslog] source-address value of fxp0 of RE1 under your re1 apply-group. > That may supersede default-address-selection. > > /Chris > > On Mon, Aug 31, 2015, at 06:40 PM, Michael Hare wrote: > > Hello- > > > > I know there have been a million threads on this topic, but I don't see > > this particular variant being discussed. > > > > To date we have managed our Juniper devices inline [via lo0, default > > address selection] instead of via fxp. However, I would like to collect > > syslogs from backup RE so I have begun connecting re0 and re1 fxp0 > > interfaces and using backup-router/inet6-backup-router via re0/re1 groups > > with the master-only flag for basic IP connectivity and monitoring. > > > > I have observed that despite lack of RPD, the backup RE still has all of > > its interfaces connected [including lo0], and we use > > default-address-selection, so backup RE sources from lo0 meaning that by > > IP address I cannot tell if the syslog is coming from master RE or backup > > RE. > > > > Other than resorting to commit scripts, I couldn't figure out a way to > > continue receiving syslogs from lo0 [via default address selection] from > > master RE but use backup RE's fxp0 address as source for backup RE > > traffic. > > > > I've read about the alternatives [logical systems, everything else in a > > vrf]. Not going down that path, I'd rather live with suboptimal if > > that's the only way. > > > > -Michael > > _______________________________________________ > > 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