Activity log for bug #1870144

Date Who What changed Old value New value Message
2020-04-01 14:06:32 Christian Ehrhardt  bug added bug
2020-04-02 06:16:13 Launchpad Janitor merge proposal linked https://code.launchpad.net/~paelzer/ubuntu/+source/chrony/+git/chrony/+merge/381562
2020-04-02 06:16:45 Christian Ehrhardt  chrony (Ubuntu): status New In Progress
2020-04-02 06:17:17 Christian Ehrhardt  summary chrony test 100-clockupdate flaky in autopkgtest autopkgtest identified fighting time services
2020-04-02 06:26:05 Christian Ehrhardt  description FYI Separated from bug 1867036 It seems to be racy/flaky. It uses a comparison on "seconds" granularity. I was adding debug to check the values it fails on. It was somewhat reproducible in the bileto ticket (again autopkgtest infrastructure). Therefore I inserted some debug code and re-ran it there. Our test environment is too fast (sometimes) and hence the flakyness: checking chronyc output OK DEBUG before 1585747433 DEBUG before 1585747433 checking system clock BAD That is checked with "lt" and here the results are ==. I'll need to discuss if there is a strict reason for these to be non-equal. Otherwise the fix might be as easy as using "-le" instead This SRU Template is meant to help the release team accepting this into focal before release. [Impact] * Over the years there were multiple approaches to adapt different time service like ntp/chrony/systemd-timesyncd to work together. The recent Debian changes (3.5-6) fixed a certain type of issues of the former solution (conflicts) but opened another issue after install until reboot where now multiple service are active. Overall this: * Fix autopkgtests to not have chrony installed (fighting with the chrony service started by the tests) * Fix autopkgtests to disable systemd-timesyncd to not fight over the clock * Reload systemd-timesyncd after chrony install to pick up the drop in config placed by chrony * Adding a new test that ensures this is detected earlier in the future in case it reoccurs (again) [Test Case] * Install chrony, the state should be: - chrony running - systemd-timesyncd inactive => That should persist over reboots => In Ubuntu that even applies to containers [Regression Potential] * I don't think it is an issue , but I could see potential regressions in the restart of systemctl-timesyncd if people made customizations to it. Therefore we ignore the RC to not break chrony install - but it would be "please fix your config" as usual. Also important is that this would not be different to the situation these users would have had after a reboot. [Other Info] * n/a ---- FYI Separated from bug 1867036 It seems to be racy/flaky. It uses a comparison on "seconds" granularity. I was adding debug to check the values it fails on. It was somewhat reproducible in the bileto ticket (again autopkgtest infrastructure). Therefore I inserted some debug code and re-ran it there. Our test environment is too fast (sometimes) and hence the flakyness:   checking chronyc output OK     DEBUG before 1585747433     DEBUG before 1585747433   checking system clock BAD That is checked with "lt" and here the results are ==. I'll need to discuss if there is a strict reason for these to be non-equal. Otherwise the fix might be as easy as using "-le" instead
2020-04-02 06:26:37 Christian Ehrhardt  description This SRU Template is meant to help the release team accepting this into focal before release. [Impact] * Over the years there were multiple approaches to adapt different time service like ntp/chrony/systemd-timesyncd to work together. The recent Debian changes (3.5-6) fixed a certain type of issues of the former solution (conflicts) but opened another issue after install until reboot where now multiple service are active. Overall this: * Fix autopkgtests to not have chrony installed (fighting with the chrony service started by the tests) * Fix autopkgtests to disable systemd-timesyncd to not fight over the clock * Reload systemd-timesyncd after chrony install to pick up the drop in config placed by chrony * Adding a new test that ensures this is detected earlier in the future in case it reoccurs (again) [Test Case] * Install chrony, the state should be: - chrony running - systemd-timesyncd inactive => That should persist over reboots => In Ubuntu that even applies to containers [Regression Potential] * I don't think it is an issue , but I could see potential regressions in the restart of systemctl-timesyncd if people made customizations to it. Therefore we ignore the RC to not break chrony install - but it would be "please fix your config" as usual. Also important is that this would not be different to the situation these users would have had after a reboot. [Other Info] * n/a ---- FYI Separated from bug 1867036 It seems to be racy/flaky. It uses a comparison on "seconds" granularity. I was adding debug to check the values it fails on. It was somewhat reproducible in the bileto ticket (again autopkgtest infrastructure). Therefore I inserted some debug code and re-ran it there. Our test environment is too fast (sometimes) and hence the flakyness:   checking chronyc output OK     DEBUG before 1585747433     DEBUG before 1585747433   checking system clock BAD That is checked with "lt" and here the results are ==. I'll need to discuss if there is a strict reason for these to be non-equal. Otherwise the fix might be as easy as using "-le" instead This SRU Template is meant to help the release team accepting this into focal before release. [Impact]  * Over the years there were multiple approaches to adapt different time    service like ntp/chrony/systemd-timesyncd to work together. The recent    Debian changes (3.5-6) fixed a certain type of issues of the former    solution (conflicts) but opened another issue after install until    reboot where now multiple service are active.  Overall this:  * Fix autopkgtests to not have chrony installed (fighting with the chrony    service started by the tests)  * Fix autopkgtests to disable systemd-timesyncd to not fight over the    clock  * Reload systemd-timesyncd after chrony install to pick up the drop in    config placed by chrony  * Adding a new test that ensures this is detected earlier in the future    in case it reoccurs (again) [Test Case]  * Install chrony, the state should be:    - chrony running    - systemd-timesyncd inactive    => That should persist over reboots    => In Ubuntu that even applies to containers * Check new (and old) autopkgtest results [Regression Potential]  * I don't think it is an issue , but I could see potential regressions in    the restart of systemctl-timesyncd if people made customizations to it.    Therefore we ignore the RC to not break chrony install - but it would    be "please fix your config" as usual. Also important is that this would    not be different to the situation these users would have had after a    reboot. [Other Info]  * n/a ---- FYI Separated from bug 1867036 It seems to be racy/flaky. It uses a comparison on "seconds" granularity. I was adding debug to check the values it fails on. It was somewhat reproducible in the bileto ticket (again autopkgtest infrastructure). Therefore I inserted some debug code and re-ran it there. Our test environment is too fast (sometimes) and hence the flakyness:   checking chronyc output OK     DEBUG before 1585747433     DEBUG before 1585747433   checking system clock BAD That is checked with "lt" and here the results are ==. I'll need to discuss if there is a strict reason for these to be non-equal. Otherwise the fix might be as easy as using "-le" instead
2020-04-03 09:00:05 Christian Ehrhardt  chrony (Ubuntu): importance Undecided High
2020-04-03 09:00:09 Christian Ehrhardt  chrony (Ubuntu): importance High Critical
2020-04-06 05:26:24 Christian Ehrhardt  description This SRU Template is meant to help the release team accepting this into focal before release. [Impact]  * Over the years there were multiple approaches to adapt different time    service like ntp/chrony/systemd-timesyncd to work together. The recent    Debian changes (3.5-6) fixed a certain type of issues of the former    solution (conflicts) but opened another issue after install until    reboot where now multiple service are active.  Overall this:  * Fix autopkgtests to not have chrony installed (fighting with the chrony    service started by the tests)  * Fix autopkgtests to disable systemd-timesyncd to not fight over the    clock  * Reload systemd-timesyncd after chrony install to pick up the drop in    config placed by chrony  * Adding a new test that ensures this is detected earlier in the future    in case it reoccurs (again) [Test Case]  * Install chrony, the state should be:    - chrony running    - systemd-timesyncd inactive    => That should persist over reboots    => In Ubuntu that even applies to containers * Check new (and old) autopkgtest results [Regression Potential]  * I don't think it is an issue , but I could see potential regressions in    the restart of systemctl-timesyncd if people made customizations to it.    Therefore we ignore the RC to not break chrony install - but it would    be "please fix your config" as usual. Also important is that this would    not be different to the situation these users would have had after a    reboot. [Other Info]  * n/a ---- FYI Separated from bug 1867036 It seems to be racy/flaky. It uses a comparison on "seconds" granularity. I was adding debug to check the values it fails on. It was somewhat reproducible in the bileto ticket (again autopkgtest infrastructure). Therefore I inserted some debug code and re-ran it there. Our test environment is too fast (sometimes) and hence the flakyness:   checking chronyc output OK     DEBUG before 1585747433     DEBUG before 1585747433   checking system clock BAD That is checked with "lt" and here the results are ==. I'll need to discuss if there is a strict reason for these to be non-equal. Otherwise the fix might be as easy as using "-le" instead This SRU Template is meant to help the release team accepting this into focal before release. [Impact] Important: The fix to avoid the services running concurrently moved to bug 1849156 - this upload will only fix the test issues now. I'll keep the test steps below as I'll want to test them with the new systemd-timesyncd then.  * Over the years there were multiple approaches to adapt different time    service like ntp/chrony/systemd-timesyncd to work together. The recent    Debian changes (3.5-6) fixed a certain type of issues of the former    solution (conflicts) but opened another issue after install until    reboot where now multiple service are active.  Overall this:  * Fix autopkgtests to not have chrony installed (fighting with the chrony    service started by the tests)  * Fix autopkgtests to disable systemd-timesyncd to not fight over the    clock  * Reload systemd-timesyncd after chrony install to pick up the drop in    config placed by chrony  * Adding a new test that ensures this is detected earlier in the future    in case it reoccurs (again) [Test Case]  * Install chrony, the state should be:    - chrony running    - systemd-timesyncd inactive    => That should persist over reboots    => In Ubuntu that even applies to containers  * Check new (and old) autopkgtest results [Regression Potential]  * I don't think it is an issue , but I could see potential regressions in    the restart of systemctl-timesyncd if people made customizations to it.    Therefore we ignore the RC to not break chrony install - but it would    be "please fix your config" as usual. Also important is that this would    not be different to the situation these users would have had after a    reboot. [Other Info]  * n/a ---- FYI Separated from bug 1867036 It seems to be racy/flaky. It uses a comparison on "seconds" granularity. I was adding debug to check the values it fails on. It was somewhat reproducible in the bileto ticket (again autopkgtest infrastructure). Therefore I inserted some debug code and re-ran it there. Our test environment is too fast (sometimes) and hence the flakyness:   checking chronyc output OK     DEBUG before 1585747433     DEBUG before 1585747433   checking system clock BAD That is checked with "lt" and here the results are ==. I'll need to discuss if there is a strict reason for these to be non-equal. Otherwise the fix might be as easy as using "-le" instead
2020-04-06 10:56:53 Launchpad Janitor chrony (Ubuntu): status In Progress Fix Released