** Description changed: [Impact] - * Some hibernated, then started instances don't restore network - connectivity keeping the instance unreachable. + * Some hibernated, then started instances don't restore network connectivity keeping the instance unreachable. + * The fix is restarting systemd-networkd on resume. [Test Case] - 0. Start a ???? instance from an encrypted EBS-backed AMI, with + 0. Start a ???? instance from an encrypted EBS-backed AMI, with hibernation enabled. - 1. Install ec2-hibinit-agent + 1. Install ec2-hibinit-agent - 2. Start a long running process on the instance, like top in screen. + 2. Start a long running process on the instance, like top in screen. - 3. Hibernate, then after it finished start the instance on EC2 console + 3. Hibernate, then after it finished start the instance on EC2 console - 4. Log in to the instance and observe top still running in screen. + 4. Log in to the instance and observe top still running in screen. - 5. Hibernate, then after it finished start the instance on EC2 console + 5. Hibernate, then after it finished start the instance on EC2 console - 6. Log in to the instance and observe top still running in screen. - (This second cycle ensures that hibernation works more than once.) + 6. Log in to the instance and observe top still running in screen. + (This second cycle ensures that hibernation works more than once.) [Regression Potential] - TODO + * Restarting systemd-networkd may cause disturbances in complex networking setups, but since the system was hibernated networking was down anyway. + * The hook in /lib/systemd/system-sleep/ is ran in parallel to other hooks in the same directory and restarting networking may break them. In Bionic the following packages use similar hooks: + + $ apt-file search /lib/systemd/system-sleep/ + atop: /lib/systemd/system-sleep/atop-pm + battery-stats: /lib/systemd/system-sleep/battery-stats + ec2-hibinit-agent: /lib/systemd/system-sleep/hibinit-agent + hdparm: /lib/systemd/system-sleep/hdparm + lizardfs-chunkserver: /lib/systemd/system-sleep/lizardfs-chunkserver + tuxonice-userui: /lib/systemd/system-sleep/tuxonice + unattended-upgrades: /lib/systemd/system-sleep/unattended-upgrades + + Only lizardfs-chunkserver may be affected because it starts lizardfs- + chunkserver.service on resume but by the description it claims to be + reliable thus a networking restart is probably tolerated, too. Also it + has ~50 popcon count in Debian which may not warrant an extensive + investigation nor adding Breaks: without being sure that it breaks.
-- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1830427 Title: Instance may loose network connectivity after resuming the 2nd time To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/ec2-hibinit-agent/+bug/1830427/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
