been started we start docker (if enabled) and then libvirt (if enabled), and only if array is Started in
non-maintenance mode. Upon array Stop, libvirt is stopped first, then docker, then network services.
The libvirt loopback image now exists on the "system" system share. Script ./dynamix.vm.manager/scripts/libvirt_init
is invoked each time libvirt is started to handle initialization of the mounted image file if necessary. Also got
rid of the old way libvirt loopback was initialized via compressed tar file.
Create/mount of docker and libvirt loopback images implemented in /usr/local/sbin/mount_image script.
The docker and libvirt loopback image files by default are located on /mnt/user/system/docker.img and
/mnt/user/system/libvirt/libvirt.img.
Default /boot/config/docker.cfg and /boot/config/domains.cfg added to flash image. This is so that docker and
libvirt start out enabled.
The /boot/config/shares/{appdata,domains,isos,system}.cfg files are created each time array Starts if they do
not exist already. This establishes defaults for these so-called "system" shares.
The "appdata", "domains", and "system" system shares have setting shareUseCache="only". These shares therefore
cannot be auto-created unless/until a cache disk/pool is available.
- Dont start container after update if it was in a stopped state before the update
- Dont try to stop a container if it is not running (was causing warnings to be logged)
- Reduced script tag generation for docker pull operations
- Correctly clean up old orphan images after updating a container (was broken before)
- Eliminated race condition when deleting both container and image
- Switched to use SweetAlert instead of jQuery UI dialogs for prompts and alerts
- Trial key consults keyserver upon boot to get time for purposes of
validating expiration. Server time in general is not affected and ntp
is still used to maintain that.
- Trial attached devices limit increased from 3 to 6.
- Support NVMe array/cache devices
- Devices coming online due to format complete immediately available
without having to stop/start array.
- Shares marked cache-only can still be created if no cache disk
exists or if not enough space left on cache disk. Same for objects
created on cache-only share.
- Numerous misc. code improvements
Known issues:
- Temperature for SAS and NVMe not obtained correctly. SCSI does not
support SMART but smartctl can still be used to get temperatures -
need to add code to do this however.
