Re: VRF Issue Since kernel 5

2 messages, 2 authors, 2019-09-12 · open the first message on its own page

Re: VRF Issue Since kernel 5

From: David Ahern <hidden>
Date: 2019-09-11 16:09:44

On 9/11/19 3:01 PM, Gowen wrote:
Hi all,

It looks like ip vrf exec checks /etc/resolv.conf (found with strace -e
trace=file sudo ip vrf exec mgmt-vrf host www.google.co.uk &>
~/straceFileOfVrfHost.txt) , but as I'm on an Azure machine using
netplan, this file isn't updated with DNS servers. I have added my DNS
server to resolv.conf and now can update the cache with "sudo ip vrf
exec sudo apt update", if I am correct (which I'm not sure about as not
really my area) then this might be affecting more than just me.

Also still not able to fix the updating cache from global VRF - which
would cause bother in prod environment to others as well so think it
would be good to get an RCA for it?

thanks for your help so far, has been really interesting.

Gareth


------------------------------------------------------------------------
*From:* Gowen [off-list ref]
*Sent:* 11 September 2019 13:48
*To:* David Ahern [off-list ref]; Alexis Bauvin
[off-list ref]; mmanning@vyatta.att-mail.com
[off-list ref]
*Cc:* netdev@vger.kernel.org [off-list ref]
*Subject:* Re: VRF Issue Since kernel 5
 
yep no problem:

Admin@NETM06:~$ sudo sysctl -a | grep l3mdev
net.ipv4.raw_l3mdev_accept = 1
net.ipv4.tcp_l3mdev_accept = 1
net.ipv4.udp_l3mdev_accept = 1


The source of the DNS issue in the vrf exec command is something to do
with networkd managing the DNS servers, I can fix it by explicitly
mentioning the DNS server:

systemd-resolve --status --no-page

<OUTPUT OMITTED>

Link 4 (mgmt-vrf)
      Current Scopes: none
       LLMNR setting: yes
MulticastDNS setting: no
      DNSSEC setting: no
    DNSSEC supported: no

Link 3 (eth1)
      Current Scopes: DNS
       LLMNR setting: yes
MulticastDNS setting: no
      DNSSEC setting: no
    DNSSEC supported: no
         DNS Servers: 10.24.65.203
                      10.24.65.204
                      10.25.65.203
                      10.25.65.204
          DNS Domain: reddog.microsoft.com

Link 2 (eth0)
      Current Scopes: DNS
       LLMNR setting: yes
MulticastDNS setting: no
      DNSSEC setting: no
    DNSSEC supported: no
         DNS Servers: 10.24.65.203
                      10.24.65.204
                      10.25.65.203
                      10.25.65.204
          DNS Domain: reddog.microsoft.com

there is no DNS server when I use ip vrf exec command (tcpdump shows
only loopback traffic when invoked without my DNS sever explicitly
entered) - odd as mgmt-vrf isnt L3 device so thought it would pick up
eth0 DNS servers?

I dont think this helps with my update cache traffic from global vrf
though on port 80
Let's back up a bit: your subject line says vrf issue since kernel 5.
Did you update / change the OS as well?

ie., the previous version that worked what is the OS and kernel version?
What is the OS and kernel version with the problem?

Re: VRF Issue Since kernel 5

From: Gowen <hidden>
Date: 2019-09-12 06:55:05

currently:

vmAdmin@NETM06:~$ uname -r
5.0.0-1018-azure

vmAdmin@NETM06:~$ cat /etc/lsb-release
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=18.04
DISTRIB_CODENAME=bionic
DISTRIB_DESCRIPTION="Ubuntu 18.04.3 LTS"


I don't keep a history of kernel versions on test but I noticed it had gone to kernel 5 and stopped working - I'm about 80% sure that happened at the same time - I'll try and dig out some logs today to see what I can find for you as Linux is fairly new to me

Gareth





From: David Ahern <redacted>

Sent: 11 September 2019 17:09

To: Gowen <redacted>; Alexis Bauvin <redacted>; mmanning@vyatta.att-mail.com <redacted>

Cc: netdev@vger.kernel.org <redacted>

Subject: Re: VRF Issue Since kernel 5

 


On 9/11/19 3:01 PM, Gowen wrote:
Hi all,
It looks like ip vrf exec checks /etc/resolv.conf (found with strace -e
trace=file sudo ip vrf exec mgmt-vrf host www.google.co.uk &>
~/straceFileOfVrfHost.txt) , but as I'm on an Azure machine using
netplan, this file isn't updated with DNS servers. I have added my DNS
server to resolv.conf and now can update the cache with "sudo ip vrf
exec sudo apt update", if I am correct (which I'm not sure about as not
really my area) then this might be affecting more than just me.
Also still not able to fix the updating cache from global VRF - which
would cause bother in prod environment to others as well so think it
would be good to get an RCA for it?
thanks for your help so far, has been really interesting.
Gareth
------------------------------------------------------------------------
*From:* Gowen [off-list ref]
*Sent:* 11 September 2019 13:48
*To:* David Ahern [off-list ref]; Alexis Bauvin
[off-list ref]; mmanning@vyatta.att-mail.com
[off-list ref]
*Cc:* netdev@vger.kernel.org [off-list ref]
*Subject:* Re: VRF Issue Since kernel 5
 
yep no problem:
Admin@NETM06:~$ sudo sysctl -a | grep l3mdev
net.ipv4.raw_l3mdev_accept = 1
net.ipv4.tcp_l3mdev_accept = 1
net.ipv4.udp_l3mdev_accept = 1
The source of the DNS issue in the vrf exec command is something to do
with networkd managing the DNS servers, I can fix it by explicitly
mentioning the DNS server:
systemd-resolve --status --no-page
<OUTPUT OMITTED>
Link 4 (mgmt-vrf)
      Current Scopes: none
       LLMNR setting: yes
MulticastDNS setting: no
      DNSSEC setting: no
    DNSSEC supported: no
Link 3 (eth1)
      Current Scopes: DNS
       LLMNR setting: yes
MulticastDNS setting: no
      DNSSEC setting: no
    DNSSEC supported: no
         DNS Servers: 10.24.65.203
                      10.24.65.204
                      10.25.65.203
                      10.25.65.204
          DNS Domain: reddog.microsoft.com
Link 2 (eth0)
      Current Scopes: DNS
       LLMNR setting: yes
MulticastDNS setting: no
      DNSSEC setting: no
    DNSSEC supported: no
         DNS Servers: 10.24.65.203
                      10.24.65.204
                      10.25.65.203
                      10.25.65.204
          DNS Domain: reddog.microsoft.com
there is no DNS server when I use ip vrf exec command (tcpdump shows
only loopback traffic when invoked without my DNS sever explicitly
entered) - odd as mgmt-vrf isnt L3 device so thought it would pick up
eth0 DNS servers?
I dont think this helps with my update cache traffic from global vrf
though on port 80


Let's back up a bit: your subject line says vrf issue since kernel 5.

Did you update / change the OS as well?



ie., the previous version that worked what is the OS and kernel version?

What is the OS and kernel version with the problem?
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help