Hello,
maybe it is a platform-specific bug... Please append this snippet to your configuration
file and gather logs:
event.recurrent(10*sec, function()
        local kluautil = require('kluautil')
        print(table_print(kluautil.list_dir(worker.control_path)))
        os.execute('ls -la ' .. worker.control_path)
end)
Every 10 seconds it will print how kresd "sees" control sockets together with
output from `ls -la` (i.e. how it is seen by OS).
Ideally we need to see debug print around time when the problem manifests. Once you gather
logs do not forget to remove these lines and restart resolver.
Thank you for getting in touch!
Petr Špaček  @  CZ.NIC
On 13. 11. 20 10:59, Petr Kyselák wrote:
  Hello, thank you for your quick response.
 root@dns-cache-2:/# ls -la /run/knot-resolver/control
 total 0
 drwxr-x--- 2 knot-resolver knot-resolver 60 Nov 13 09:28 .
 drwxr-x--- 3 knot-resolver knot-resolver 60 Jul 19 17:52 ..
 srwxr-xr-x 1 knot-resolver knot-resolver  0 Nov 13 09:28 1
 Yes, systemd. No, I do not explicitly bind in the configuration file. 
 # systemctl start kresd(a)1.service
 pá 13. 11. 2020 v 10:47 odesílatel Tomas Krizek <tomas.krizek(a)nic.cz
<mailto:tomas.krizek@nic.cz>> napsal:
     Hi,
     it's probably caused by some weird control socket names and our new
     implementation of map().
     Could you list the directory with control sockets?
     # ls -la /run/knot-resolver/control
     Also, how do you start the kresd services? Do you use systemd? If so,
     what are the instance names? Or do you perhaps explicitly bind to
     control sockets in your configuration file?
     Thanks.
     On 13/11/2020 10.33, Petr Kyselák wrote:
  Hello,
 recently we upgraded from 5.1 to 5.2 few servers (CentOS7 and Raspbian) and
 all seems to be working fine, but I can see on (yes I know, not
 recommend and supported) rasbian issue with metrics and in the log I can
 issue this error:
 Nov 13 10:21:33 dns-cache-2 kresd[15232]: map() error while connecting to
 control socket /run/knot-resolver/control/H#003: socket:connect: No such
 file or directory (ignoring this socket)
 Nov 13 10:21:33 dns-cache-2 kresd[15232]: map() error while connecting to
 control socket /run/knot-resolver/control/H: socket:connect: No such file
 or directory (ignoring this socket)
 Nov 13 10:21:33 dns-cache-2 kresd[15232]: map() error while connecting to
 control socket /run/knot-resolver/control/: socket:connect: Connection
 refused (ignoring this socket)
 the error is triggered  by opening "ip:8453/metrics" and it show almost
 empty response:
 # TYPE resolver_latency histogram
 resolver_latency_count 0.000000
 resolver_latency_sum 0.000000
 root@dns-cache-2:/# apt -qq list knot* --installed
 knot-resolver-module-http/unknown,now 5.2.0-1 all [installed]
 knot-resolver-release/unknown,now 1.7-1 all [installed]
 knot-resolver/unknown,now 5.2.0-1 armhf [installed]
 root@dns-cache-2:/#
 root@dns-cache-2:/# uname -a
 Linux dns-cache-2 4.19.97-v7l+ #1294 SMP Thu Jan 30 13:21:14 GMT 2020
 armv7l GNU/Linux
 Could it be a bug or something related to mix armv7l and armhf? I know of
 the limitation as we discussed this setup on raspberry some time ago. I
 just want to share my experience as it can be useful and if you will have
 any tip I will appreciate it.
 
     --
     Tomas Krizek
     PGP: 4A8B A48C 2AED 933B D495  C509 A1FB A5F7 EF8C 4869