Hello, On 02. 09. 19 11:46, Milan Jeskynka Kazatel wrote: Here are multiple options: a) If you do not require zero-downtime then you can simply restart kresd@1 instance: systemctl restart kresd@1 b) If you require zero-downtime and have only single kresd instance you can do this: systemctl start kresd@2 systemctl restart kresd@1 systemctl stop kresd@2 Please note the numbers at the end! These commands start second (temporary) Knot Resolver instance with new config and then restart the old instance with old config. During restart the second instance will handle incomming DNS queries. After this exercise the second instance is not needed and can be stopped. c) If you require zero-downtime and have more than one kresd instance you can simply restart instances one after the other: systemctl restart kresd@1 systemctl restart kresd@2 systemctl restart kresd@3 ... In general we recommend having at least two instances running because it allows to handle fatal errors (like crash) by automatic restart - and clients will likely not notice anything. I hope it helps. -- Petr Špaček @ CZ.NIC

Hello Vladimir,  if I understand your hint correctly, then I should change in the config file a variable from hints.add_hosts([path]) to hints.config([path]) because it does the same things + reload hints. "Clear any configured hints, and optionally load a hosts-like file as in  hints.add_hosts(path). (Root hints are not touched.)" Could you please more specify the "command" how to in the script attach it to the control socket? You mention a echo. Anyway, the hint from Petr Spacek sounds good for my case. *** If you require zero-downtime and have only single kresd instance you  can do this:  systemctl start kresd@2  systemctl restart kresd@1  systemctl stop kresd@2  Please note the numbers at the end!  *** I need to know what option is more suggested and supports daemon-reload in the case when the option is missing.  Why is not exists so simply command like "systemctl reload kresd@1" Thank you for your time and advice. -- Smil Milan Jeskyňka Kazatel ---------- Původní e-mail ---------- Od: Vladimír Čunát <vladimir.cunat(a)nic.cz> Komu: Milan Jeskynka Kazatel <KazatelM(a)seznam.cz> Datum: 2. 9. 2019 12:52:47 Předmět: Re: [knot-resolver-users] Knot Resolver - module hints - reload? "Hello. On 9/2/19 11:46 AM, Milan Jeskynka Kazatel wrote: In this particular case there's also a command that clears the hints and (optionally) loads a file: https://knot-resolver.readthedocs.io/en/stable/modules.html#c.hints.config  You can run commands by attaching to the control socket, e.g. echo "command" | nc -U /run/knot-resolver/control@1 --Vladimir "

Hello Vladimir, many thanks for your explanation regarding your meaning of command with using control@1. But I'm failed. Unfortunately, I'm using now version Knot DNS Resolver, version 2.3.0 (I really want to upgrade it later) trough a systemd with module HTTP on Centos 7. For me, the kresd does not create any control@1 nowhere, then I´m not able to use **echo "hints.config('/etc/hosts')" | sudo nc -U /run/knot-resolver/ control@1** And unfortunately, my second option with two instances of kresd@1 and kresd@ 2, mentioned by Petr Špaček @ CZ.NIC,  failed because my HTTP address and port are already in use. Any other option how silently do a "reload" hints. in the case when is a record in this file changed? thanks,  -- Smil Milan Jeskyňka Kazatel ---------- Původní e-mail ---------- Od: Vladimír Čunát <vladimir.cunat(a)nic.cz> Komu: Milan Jeskynka Kazatel <KazatelM(a)seznam.cz> Datum: 2. 9. 2019 15:10:44 Předmět: Re: [knot-resolver-users] Knot Resolver - module hints - reload? "On 9/2/19 3:00 PM, Milan Jeskynka Kazatel wrote: No, changing the config isn't needed (and won't help).  The hints.config command clears all hints and loads new ones from a file.  Let me write a more explicit example doing the reload: echo "hints.config('/etc/hosts')" | sudo nc -U /run/knot-resolver/control@1 --Vladimir "

Hello Vladimir, many thanks, the  /var/cache/knot-resolver/tty/$PID works fine. Regards. -- Smil Milan Jeskyňka Kazatel ---------- Původní e-mail ---------- Od: Vladimír Čunát <vladimir.cunat(a)nic.cz> Komu: Milan Jeskynka Kazatel <KazatelM(a)seznam.cz> Datum: 5. 9. 2019 12:45:12 Předmět: Re: [knot-resolver-users] Knot Resolver - module hints - reload? "On 9/5/19 11:37 AM, Milan Jeskynka Kazatel wrote: Oh, for CentOS the packaging is a bit different than elsewhere.  I think the control socket is located in /var/cache/knot-resolver/tty/$PID - at least in the current 4.2.0 - anyway, if you find that file, it should work the same even for 2.3.0. I think this particular problem was solved in 4.0.0 Restarting the service is probably best in your current situation.  The cache is kept, so it should be cheap (and easy!). --Vladimir "