In MX and PTX Series routers, the Simple Network Management Protocol (SNMP) may run into a defect where snmpwalk dot3adAggPortPartnerOperPort or dot3adAggPortAttachedAggID returns no output.
This article suggests a workaround and indicates the Junos OS releases in which the issue has been resolved.
snmpwalk dot3adAggPortPartnerOperPort or dot3adAggPortAttachedAggID returns no output as shown in the example below:
Non-Working One
{MASTER}
lab@router> show snmp mib walk ifDescr | match ae308
ifDescr.1227 = ae308
ifDescr.1853 = ae308.3808
ifDescr.1854 = ae308.32767
{MASTER}
lab@router> show snmp mib walk dot3adAggPortAttachedAggID | match 1227
<<<<<<<<<< no output
{MASTER}
lab@router> show lacp interfaces ae308
Aggregated interface: ae308
LACP state: Role Exp Def Dist Col Syn Aggr Timeout Activity
xe-7/1/5 Actor No No Yes Yes Yes Yes Fast Active
xe-7/1/5 Partner No No Yes Yes Yes Yes Fast Active
LACP protocol: Receive State Transmit State Mux State
xe-7/1/5 Current Fast periodic Collecting distributing
Working One
{MASTER}
lab@router> show snmp mib walk ifDescr | match ae308
ifDescr.1231 = ae308
ifDescr.1852 = ae308.32767
ifDescr.1853 = ae308.3708
{MASTER}
lab@router> show snmp mib walk dot3adAggPortAttachedAggID | match 1231
dot3adAggPortAttachedAggID.1854 = 1231
dot3adAggPortAttachedAggID.1855 = 1231
{MASTER}
lab@router> show lacp interfaces ae308
Feb 20 03:46:10
Aggregated interface: ae308
LACP state: Role Exp Def Dist Col Syn Aggr Timeout Activity
xe-7/1/5 Actor No No Yes Yes Yes Yes Fast Active
xe-7/1/5 Partner No No Yes Yes Yes Yes Fast Active
LACP protocol: Receive State Transmit State Mux State
xe-7/1/5 Current Fast periodic Collecting distributing
Lag children interfaces are learned via a stats request sent to the parent ae interface because the stats response for a lag parent interface contains stats for all individual lag children interfaces. This stats response helps mib2d to detect all the children interfaces associated with a particular lag ae port.
In this case (for example, router reboot), the FPC might be slow to populate all the children, resulting in the initial stats request timing out in the kernel -> PFC layer. Hence, mib2d receives an empty stats response and is not able to learn the lag children interfaces. This state continues to exist until a change happens on the ae interface, which triggers another request to refresh/repopulate the lag children interfaces or until mib2d is restarted, which will also trigger a stats request for relearning the ae parent interface.
As per PR1220988, the issue has been resolved in Junos OS releases ​16.1R3-S4, 16.1R4-S3, 16.1R5, 16.2R3, and 17.1R1.
Meanwhile, a workaround is to restart mib2d
with the following command:
restart mib-process​