+++ This bug was initially created as a clone of Red HatBug #1488218 +++

Description of problem:
kill -HUP PID_OF_NAMED doesn't return error if named.conf has wrong configuration 

Version-Release number of selected component (if applicable):

bind-9.9.4-51

How reproducible:


Steps to Reproduce:
Reproduce steps:
- add include statement in bind configuration
include "myextraconfig";
- do NOT create the "myextraconfig" file
- confirm broken bind configuration with "named-checkconf"
- run "systemctl reload named"
- or run "kill -HUP <Named PID>"
- check exit code "echo $?" --- It returns 0 instead of 1 

Actual results:
exit code 0

Expected results:
It should return 1

--- Additional comment from Tomáš Hozza 🤓 on 2017-09-05 09:18:59 CEST ---

Hello.

"kill" just sends signal to a process. It does not checks what the process actually does. Therefore the fact that kill returns 0 is expected and not a bug.

--- Additional comment from Ramesh Sahoo on 2017-09-13 18:50:42 CEST ---


If someone executing reload command it must return something i.e. either success or error. If kill -HUL can't fulfill this then we shouldn't use this at all and if we use this my suggestion to replace the below parameter in reload which definitely returns something. 

named-checkconf > /dev/null 2>&1 && { rndc reload > /dev/null 2>&1 || /bin/kill -HUP $MAINPID; } || exit 1

--- Additional comment from Petr Menšík on 2017-09-15 12:19:16 CEST ---

Hi Ramesh.

Current state is a bit unfortunate. rndc reload command returns error when configuration error is found. Rdnc configuration is enabled by default, but kill is used in case rndc cannot connect to the server. It cannot tell difference between reload failed because there is configuration error and failed to connect to the server at all. If you are sure rndc works and if not, it is an error, you can remove " || /bin/kill -HUP $MAINPID" from ExecReload.

Extension of systemd service file can be used for this.
Create file /etc/systemd/system/named.service.d/reload.conf with contents:

; Simply rely on rndc to be properly configured to communicate with local server
[Service]
ExecReload=
ExecReload=/usr/sbin/rndc reload

If used like that, any error in reload would be reported by systemctl. Some users may not want to use rndc at all.
Or use named-checkconf the same way like on start. If named-checkconf fails, no reload command or signal would be sent.

; More compatible solution
[Service]
ExecReload=
ExecReload=/bin/bash -c 'if [ ! "$DISABLE_ZONE_CHECKING" == "yes" ]; then /usr/sbin/named-checkconf -z /etc/named.conf; else echo "Checking of zone files is disabled"; fi'
ExecReload=/bin/sh -c '/usr/sbin/rndc reload > /dev/null 2>&1 || /bin/kill -HUP $MAINPID'

run "systemctl daemon-reload" to load extension into running systemd.

Some runtime errors are not catched by named-checkconf however. It would be nice if rndc reload could return different code for failed operation reported by the server and failed connection to the server. It would be able use kill -HUP only when connection to the server failed, not when server already indicated error on reload. No such feature exist even in lastest development releases however.

--- Additional comment from Petr Menšík on 2019-07-30 11:34:48 CEST ---

rndc null can be used to verify connection to named without doing anything actually. Makes possible to distinguish whether connection to named failed in case of disabled control channel in configuration.

; Compatible solution reporting errors on rndc reload failure
[Service]
ExecReload=
ExecReload=/bin/bash -c 'if [ ! "$DISABLE_ZONE_CHECKING" == "yes" ]; then /usr/sbin/named-checkconf -z /etc/named.conf; else echo "Checking of zone files is disabled"; fi'
ExecReload=/bin/sh -c 'if /usr/sbin/rndc null > /dev/null 2>&1 ; then /usr/sbin/rndc reload; else /bin/kill -HUP $MAINPID; fi'

This way, rndc null errors are ignored and kill is used instead. If null succeed, reload is requested. Failures in that are reported.

--- Additional comment from Petr Menšík on 2019-08-08 11:22:59 CEST ---

Oh, proposed solution with if would not work, because of systemd bug [1]. It makes reload terminate the service, which is not desired at all. Return code has to be ignored until this is fixed. It makes impossible to report failure in reload in different way than just log message. RHEL 8 bug #1735787 fixed it.

Until that is fixed, it may just log error into log file:
# /etc/systemd/system/named.service.d/reload.conf
; vim:set ft=systemd:
[Service]
ExecReload=
ExecReload=-/bin/sh -c 'if /usr/sbin/rndc null > /dev/null 2>&1; then /usr/sbin/rndc reload || true; else /bin/kill -HUP $MAINPID; fi'


1. https://github.com/systemd/systemd/pull/13098