I filed an MP for impish and and MR for Debian at https://salsa.debian.org/dns-team/bind9/-/merge_requests/17 with the missing commit and a regression autopkgtest test case.
I will start working on SRUs after I get reviews on those. Note that there may be an issue with testing the changes using nginx in impish, which is being addressed here: https://salsa.debian.org/dns- team/bind9/-/merge_requests/17/diffs#146a89a78e4846ec68c44d1e006f142585418863_0_81 The reason I am mentioning nginx in particular is that this is the package that led the reporter to file this bug against bind9. If you are not interested in the resolution for nginx, you can skip the rest of this comment. The nginx package in impish has an ExecStartPre= directive in it's systemd unit file. While testing the patch above with the provided autopkgtest test case (without the sed command to remove the ExecStartPre= directive) I was facing a situation where the test would succeed in some of the runs, and fail in other runs. The issue did not happen for focal, which does not contain the ExecStartPre= directive in the nginx unit file. In the failing cases, checking the systemd service status (or logs) for nginx would contain errors on failed configuration file checks (which is what the ExecStartPre= directives does in the impish nginx package), hindering the service start up process: ``` Jun 18 18:37:41 test-bind9-repr-script-impish systemd[1]: Starting A high performance web server and a reverse proxy server... Jun 18 18:37:41 test-bind9-repr-script-impish nginx[285]: nginx: [emerg] host not found in upstream "bind9test.debian.dev" in /etc/nginx/sites-enabled/default:4 Jun 18 18:37:41 test-bind9-repr-script-impish nginx[285]: nginx: configuration file /etc/nginx/nginx.conf test failed Jun 18 18:37:41 test-bind9-repr-script-impish systemd[1]: nginx.service: Control process exited, code=exited, status=1/FAILURE Jun 18 18:37:41 test-bind9-repr-script-impish systemd[1]: nginx.service: Failed with result 'exit-code'. Jun 18 18:37:41 test-bind9-repr-script-impish systemd[1]: Failed to start A high performance web server and a reverse proxy server. ``` Sometimes, disabling the nginx service, enabling it back, and rebooting the system was the only way I could get the issue to manifest itself (which would still be an intermittent issue). Here is a short reproducer for the issue I am observing. un-commenting the Wants bit does the trick and get everything to start in the expected order. ``` #!/bin/bash set -xe MAIN_UNIT=$HOME/.config/systemd/user/reproducer-main.service DEPENDENCY_UNIT=$HOME/.config/systemd/user/reproducer-dependency.service TMP_LOGFILE=$(tempfile) cat << EOF > $MAIN_UNIT [Unit] Description=testing stuff After=reproducer-dependency.service # Wants=reproducer-dependency.service [Service] Type=forking ExecStartPre=sh -c 'echo I should be msg 2 >> $TMP_LOGFILE' ExecStart=sh -c 'echo I should be msg 3 >> $TMP_LOGFILE' [Install] WantedBy=default.target EOF cat << EOF > $DEPENDENCY_UNIT [Unit] Description=This should be triggered BEFORE everything else [Service] Type=forking ExecStart=sh -c 'echo I should be msg 1 >> $TMP_LOGFILE' [Install] WantedBy=default.target EOF echo 'Starting order with the unit containing the After= directive first' >> $TMP_LOGFILE systemctl --user start reproducer-main.service reproducer-dependency.service systemctl --user stop reproducer-main.service reproducer-dependency.service echo echo 'Starting order with the unit containing the After= directive last' >> $TMP_LOGFILE systemctl --user start reproducer-dependency.service reproducer-main.service systemctl --user stop reproducer-dependency.service reproducer-main.service cat $TMP_LOGFILE rm -f $TMP_LOGFILE $MAIN_UNIT $DEPENDENCY_UNIT ``` This is the output I get from it: ``` Starting order with the unit containing the After= directive first I should be msg 2 I should be msg 1 I should be msg 3 Starting order with the unit containing the After= directive last I should be msg 1 I should be msg 2 I should be msg 3 ``` Is this just the expected systemd behaviour (am I missing something)? I could not find anything in the docs on how ExecStartPre= directives relate to the Atfter= ones. The final consequence here is that, although the proposed patch fixes the issue, it does not address the user request on how nginx and bind9 services interact with each other. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1899902 Title: systemd unit service file does not wait for bind9 to be ready To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/bind9/+bug/1899902/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
