Public bug reported: We have some build scripts that customize a released PV instance-store image available on AWS, and build an AMI for our auto scaling group. However, the latest AMI builds extremely slowly and can't be bundled. It freezes up while trying to create the file to serve as the snapshot of the system.
My investigation points to an issue with the kernel. If I boot ami- 990d24f3 [wily 20160204, amd64 PV instance-store] in us-east-1 and check top, an obvious problem is already apparent. There's a kworker process hogging CPU. I'm seeing about 150 each of timer interrupts and hypervisor calls per second on the system in this case. I've tried customizing an older AMI, but as part of the build, we install the kernel update (from 4.2.0-25-generic) and the resulting image exhibits the same high CPU usage due to kworker, which will naturally fail our health check if it were to be released. I have also checked the image's HVM sibling, ami-291d3443, which does not exhibit the problem. This is observed on `m3.medium` instances in VPC, in at least two availability zones within the us-east-1 region. Please let me know if you need additional information. ** Affects: linux-lts-wily (Ubuntu) Importance: Undecided Status: New -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1544094 Title: kworker 100% CPU usage on Xen PV (AWS) To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux-lts-wily/+bug/1544094/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs