[Dnsmasq-discuss] DNSMasq with high Send/Receive buffers

[Felipe Polanco]

Hello,

We are using dnsmasq as DNS server for our public VPN server and today we
found a rare issue where DNSMasq filled the entire send buffer of its UDP
socket.

dnsmasq.log

Aug 24 15:52:12 dnsmasq[5363]: failed to send packet: Resource temporarily
unavailable
Aug 24 15:52:12 dnsmasq[5363]: cached www.google.com is 216.58.223.196
Aug 24 15:52:12 dnsmasq[5363]: query[A] www.google.com from 105.112.28.230
Aug 24 15:52:12 dnsmasq[5363]: failed to send packet: Resource temporarily
unavailable

# netstat -tulpana
Active Internet connections (servers and established)
Proto Recv-Q Send-Q Local Address           Foreign Address         State
    PID/Program name
tcp        0      0 0.0.0.0:111             0.0.0.0:*               LISTEN
     1/systemd
tcp        0      0 0.0.0.0:53              0.0.0.0:*               LISTEN
     5363/dnsmasq
tcp        0      0 0.0.0.0:22              0.0.0.0:*               LISTEN
     3977/sshd
<redacted>
tcp6       0      0 :::111                  :::*                    LISTEN
     1/systemd
tcp6       0      0 :::53                   :::*                    LISTEN
     5363/dnsmasq
tcp6       0      0 :::22                   :::*                    LISTEN
     3977/sshd
udp    14976 213504 0.0.0.0:53              0.0.0.0:*
    5363/dnsmasq

Low CPU usage, RAM is ok and disk is 38% used.

DNS resolution became very slow and we saw using tcpdump that maybe 1 out
of 20 requests were answered.

nslookup to 8.8.8.8 works fine.

We restarted dnsmasq service and the issue went away immediately and the Qs
were back to 0 as expected.

I understand these Qs may fill up if there is no acknowledgment from the
peer while using TCP sockets but these are UDP and don't require ACK.

What could be the cause of this?

Thanks,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.thekelleys.org.uk/pipermail/dnsmasq-discuss/attachments/20220824/d58c1934/attachment.htm>