Page MenuHomeVyOS Platform

BGP FQDN capability has improper hostname after new image install
Open, Requires assessmentPublicBUG

Description

The VyOS BGP daemon (FRR?) sends the BGP FQDN capability with the host's hostname and domain. (I'll be filing another bug on how the domain is being handled).

If a new image is installed ("add system image ...") and the system is then rebooted to make that image active, when BGP comes up it sends "debian" as the hostname in this BGP FQDN. Example as received by another BGP stack of BGP coming up after a reboot to activate a new image:

2020-02-01T00:11:32.291510Z [N] Listening
2020-02-01T00:11:32.354160Z [N] «0» 0 New connection from 192.168.64.1:179
2020-02-01T00:11:32.523416Z [N] «1» 0 BGP OPEN ASN=65001 ID=192.168.79.1 Hold-Time=180 OPT=[CAP=[MPBGP=IP/unicast];CAP=[Route-Refresh-Cisco];CAP=[Route-Refresh];CAP=[ASN32=65001];CAP=[Code=69 Len=4];CAP=[FQDN=debian,];CAP=[Graceful-Restart=120secs]]

Note the [FQDN=debian,] in that output.

If the router is rebooted again, it populates the hostname portion correctly:

2020-02-01T00:14:35.492146Z [N] «11» 1 New connection from 192.168.64.1:179
2020-02-01T00:14:35.603218Z [N] «12» 1 BGP OPEN ASN=65001 ID=192.168.79.1 Hold-Time=180 OPT=[CAP=[MPBGP=IP/unicast];CAP=[Route-Refresh-Cisco];CAP=[Route-Refresh];CAP=[ASN32=65001];CAP=[Code=69 Len=4];CAP=[FQDN=fw1.den1,];CAP=[Graceful-Restart=RESTART 120secs]]

This behavior was observed on the rolling release VyOS 1.3-rolling-202001311637

My guess is that BGP is started before the system hostname is fully configured on first boot.

Details

Difficulty level
Unknown (require assessment)
Version
1.3-rolling-202001311637
Why the issue appeared?
Will be filled on close
Is it a breaking change?
Unspecified (possibly destroys the router)

Event Timeline

jmaslak created this task.Feb 1 2020, 1:02 AM
jmaslak updated the task description. (Show Details)