I took me some time to figure out what was happening. Looking back, checking DNS server should have been one of the first things to do. Seems like first DNS server provided by my provider has been down. That meant that every DNS query timed out and then went to the second DNS which got me my response. For some reason ping did DNS query before every new package being sent. That explains its weird behaviour.
This problem got me to finally install local caching dns. I was thinking about doing it before, but I never got around to do it until now. I always thought it’s gonna be a few-hour nightmare. Now I blame my previous experience with bind 😀 For simple local caching bind would be overkill, so I chose dnsmasq. Using it was as simple as installing, running dnsmasq and executing
$ echo 'nameserver 127.0.0.1' > /etc/resolv.conf.head
From that point on every resolv.conf file generated by dhcpcd will have my local DNS as first DNS server to try. For this time you can add it there manually. Then you can verify your setup works by running following command twice in a row:
$ dig randomserver.com
First execution should have Query time: XX msec with XX being few tens of miliseconds. Query time for second run should be zero or very close to zero.
Congratulations. You have your very own caching server. Who knows…maybe you’ll even notice some improvements in your network connection