No, you don't have a dialer profile, you have legacy dialer.
Try my suggestion.
Note that virtual profiles trump legacy dialer but dialer profiles
trump virtual profiles. You don't want/need a virtual profile here,
so go with dialer profiles.
Aaron
---
~ In article <>,
(Aaron Leonard) writes:
~ | I'd get rid of all the virtual-template stuff and convert to
~ | using a dialer profile.
~
~ I don't understand your suggestion. What I have now is a dialer profile. It
~ doesn't work quite right (as described). I tried the various virtual templates
~ to improve the situation. As noted, they cause other issues. Can you tell me
~ what "multilink virtual-template" applies to exactly? And what about the route
~ re-pointing question?
~
~ | I.e. change the physical serial
~ | interfaces to do "dialer pool-member 1" and then build a
~ | dialer interface which uses "dialer pool 1".
~
~ Are you saying that grouping physical interfaces under a dialer interface
~ with the dialer pool-member command will produce different results from those
~ of using the dialer rotary-group command? What exactly is the difference?
~
~ | Aaron
~ |
~ | ---
~ |
~ | ~ I am using the following configuration for basic ppp multilink operation
~ | ~ on two synchronous serial dial lines (IOS 12.1(20) on a 2500 series):
~ | ~
~ | ~ interface Serial0
~ | ~ bandwidth 56
~ | ~ no ip address
~ | ~ encapsulation ppp
~ | ~ no ip mroute-cache
~ | ~ dialer in-band
~ | ~ dialer rotary-group 0
~ | ~ no fair-queue
~ | ~ pulse-time 1
~ | ~ compress stac
~ | ~ no cdp enable
~ | ~ !
~ | ~ interface Serial1
~ | ~ bandwidth 56
~ | ~ no ip address
~ | ~ encapsulation ppp
~ | ~ no ip mroute-cache
~ | ~ dialer in-band
~ | ~ dialer rotary-group 0
~ | ~ no fair-queue
~ | ~ pulse-time 1
~ | ~ compress stac
~ | ~ no cdp enable
~ | ~ !
~ | ~ interface Dialer0
~ | ~ ip address 1.2.3.4 255.255.255.128
~ | ~ ip access-group 100 in
~ | ~ ip accounting output-packets
~ | ~ ip pim dense-mode
~ | ~ encapsulation ppp
~ | ~ no ip mroute-cache
~ | ~ dialer in-band
~ | ~ dialer idle-timeout 2147483
~ | ~ dialer string 5551234
~ | ~ dialer-group 1
~ | ~ compress stac
~ | ~ no cdp enable
~ | ~ ppp chap hostname myname
~ | ~ ppp chap password mypassword
~ | ~ ppp multilink
~ | ~
~ | ~ This works fine (note that I left out the load threshold command so the
~ | ~ sample will bring up only one link) as far as the actual routing, though
~ | ~ a few things aren't quite right. The bundle virtual-access interface is
~ | ~ cloned from Dialer0, but the clone is flawed. At some level the system
~ | ~ believes that ip processing is disabled on the virtual-access interface
~ | ~ and thus "show ip int" cannot be used to look at, e.g., the PPP peer address.
~ | ~ Perhaps for similar reasons compression statistics are not available for
~ | ~ the virtual-access bundle interface with "show comp".
~ | ~
~ | ~ I thought that perhaps the virtual-access bundle interface was not cloning
~ | ~ the ip address from the dialer interface in order to avoid a duplicate, but
~ | ~ setting Dialer0 up as unnumbered made no difference.
~ | ~
~ | ~ I created a virtual-template with the same parameters as the dialer interface
~ | ~ and assigned it with "multilink virtual-template x". Interestingly, this
~ | ~ had no effect whatsoever, with debugging confirming that the virtual-access
~ | ~ bundle interface was still being cloned from the dialer interface. What
~ | ~ exactly does "multilink virtual-template" do? Is it perhaps for assigning
~ | ~ a template to projected links rather than bundles?
~ | ~
~ | ~ Next I assigned the virtual-template with "virtual-profile virtual-template x".
~ | ~ This worked in the sense that the virtual-access bundle interface was then
~ | ~ cloned from my template rather than from the dialer interface. Moreover,
~ | ~ "show ip int" and "show comp" did the right things. Unfortunately, the virtual
~ | ~ profile introduced a new problem in that the dialer interface was unwilling to
~ | ~ route packets from routes pointed directly to it (by, e.g., ip route 0.0.0.0
~ | ~ 0.0.0.0 Dialer0), giving an error message like "vp up dropping non-broadcast
~ | ~ packet" (I forget the exact wording). Now I can fix the latter problem by
~ | ~ making the routes point to the peer address rather than the interface, but
~ | ~ I wonder if there is a better way. Also, it isn't clear which (if any) routes
~ | ~ are adjusted to point to the virtual-access bundle interface. Obviously my
~ | ~ static default route was not, but at some level the route that routed the
~ | ~ packet that triggered the dial-on-demand event in the first place has to be
~ | ~ redirected. Perhaps only the route-to-host that is installed when IPCP comes
~ | ~ up is redirected even though the it still shows up as pointing to "Dialer0"
~ | ~ in the table? Is there any detailed documentation on this? Or about the
~ | ~ exact semantics of the various commands that apply virtual templates? The
~ | ~ documentation mentions that it is critical to understand which templates are
~ | ~ used for what, but it doesn't do a very good job explaining this.

~ | ~
~ | ~ Dan Lanciani
~ | ~ ddl@danlan.*com
~