Faulty data request

[tobpip]

Hi,

i have a problem with pybgpstrem (v2.0.2).

---

1. Problem
   When i use it to collect data than i get a error message. An excerpt is visible here.
   
   I have this problem when i try to download data after the time step 1590019200 (unix).

---

2. Problem
   When i use the "Any" filter after the time step 1584489600 (unix) i get BGP message with the prefix 0.0.0.0/0
   Before this time step i don't get such messages.

---

I did the test with different prefixes. But there was no difference in the reaction of the programme.

[vkotronis]

We are encountering the same problem on the monitors of the ARTEMIS open-source on-premise tool: https://github.com/FORTH-ICS-INSPIRE/artemis/tree/master

We are using the latest version of the libbgpstream container (2.2.0) https://github.com/slowr/artemis-base-images/blob/master/monitor/Dockerfile#L1

Our implementation (in Python) is packed here: https://github.com/FORTH-ICS-INSPIRE/artemis/blob/master/monitor-services/bgpstreamlivetap/core/bgpstreamlive.py#L369

Like @tobpip reported, we recently started getting this error that overflows the logs:

bgpstreamlivetap_1   | WARN: INVALID_MSG: Invalid prefix in MP_(UN)REACH_NLRI (parsebgp_bgp_update_mp_reach.c:110)

Is there a way to address this? Can it be suppressed in case it stems from an upstream collector? We do not see this by directly using the kafka RV stream, however all clients that use the classic Python-based client are affected (no errors, but warning overflows).

Thanks in advance for looking into this!

[racompton]

Same here. Commenting so I'm notified when this is resolved.

[alistairking]

I suspect this might be corrupted data generated by RV collectors running a version of FRR that mis-handled ADD-PATH attributes.

Can you narrow the problem down to specific collectors/time ranges?