Determine approach for unclean log disk mounting

During our gitaly rebuild tests we ran into some issues during the initial log and data disk mount: #1398 (comment 736105759).

This resulted in /var/log/syslog no longer receiving updates. Possibly because the logrotate failed, and file descriptors remaining open to the root filesystem.

Workaround

When performing a rebuild without a packer-based image, we get lucky. Because we have a kernel upgrade during the first boot, we perform a reboot, and after this reboot, things are in a good state.

We do not see that reboot with the packer-based image, because we're already on the latest kernel. However, we can simulate that behaviour by forcing a reboot.

Proposed patch to the bootstrap script:

• https://ops.gitlab.net/gitlab-com/gl-infra/terraform-modules/google/bootstrap/-/merge_requests/27

Proper fix

We could also spend some time digging into why the re-mounting is not working properly.