[j-nsp] 'lib_ifbd->ifbd_curr_num_macs >= adjust_count' failed

Lukasz Trabinski lukasz at trabinski.net
Mon Oct 21 08:25:26 EDT 2013


Hi

Strange situation on MX960 (Junos 10.4S12) with following scenario:

About 250 customers with static mac addresses and no-mac-learning options 
putted into

routing-instances XXX bridge-domains XXX

with following configuration:

interface xe-xx/x/x.xxx {
     interface-mac-limit {
         1;
         packet-action drop;
     }
     static-mac xx:xx:xx:xx:xx:x;
     no-mac-learning;


After flap (up/down) of the 10Gbit/s port

we got:

Oct 17 09:43:55  xx-xx_re0 fpc0 ICHIP(0)_REG_ERR:5 Wi seg ucode 
discards in fabric stream 40 pfe_id 20"

and some traffic to some random ports was dropped.

After that, we decided to deactivate and activate some static mac address 
and no-mac-learning

deactivate routing-instances XXX bridge-domains XXX bridge-options 
interface  aex.xx static-mac xx:xx:xx:xx:xx:x

commit

activate routing-instances XXX bridge-domains XXX bridge-options
interface  aex.xx static-mac xx:xx:xx:xx:xx:x

commit

After last commit, we got message from syslog:

xx-xx_re0 l2ald[86898]: 
../../../../src/junos/usr.sbin/l2ald/l2ald_mlrn_ifbd.h:175: insist 
'lib_ifbd->ifbd_curr_num_macs >= adjust_count' failed


and core files was dumped into /var/tmp/


Solution of this problems was:

deactivate routing-instances XXX bridge-domains XX bridge-options
commit
activate routing-instances XXX bridge-domains XX bridge-options
commit

Question is, is that software or hardware bug? Why flap of one TenGig port 
caused problem with traffic beetween other ports?



Lukasz


More information about the juniper-nsp mailing list