You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We lack a clean way for automatic, unattended installs which are required if a large number of machines are to be shipped with Seattle "preinstalled". Currently, this would require one to manually install on each machine.
It does not suffice to take a Seattle directory from an install on one machine and copy it over to another: During installation, the machine is benchmarked and VM sizes are determined, unique node keys are generated, the node manager and software updater are registered to start on boot, and both are started. Only then, the machine becomes accessible to the project. Ergo, you can't just add an existing install directory to a disk image you will later ship with each machine.
A clean way of engineering this would probably be to add a Seattle directory including the desired vessselinfo file to the disk image, and include an automatic installation rc script. That script would call the Seattle installer, and remove itself once the installer succeeded. After that, the installer will have registered the node manager and software updater to start at boot.
In the meantime, we can make the existing Seattle installation script to be run at boot in such a scenario. It will complain if it finds Seattle already installed, but not hurt otherwise.
The text was updated successfully, but these errors were encountered:
We lack a clean way for automatic, unattended installs which are required if a large number of machines are to be shipped with Seattle "preinstalled". Currently, this would require one to manually install on each machine.
It does not suffice to take a Seattle directory from an install on one machine and copy it over to another: During installation, the machine is benchmarked and VM sizes are determined, unique node keys are generated, the node manager and software updater are registered to start on boot, and both are started. Only then, the machine becomes accessible to the project. Ergo, you can't just add an existing install directory to a disk image you will later ship with each machine.
A clean way of engineering this would probably be to add a Seattle directory including the desired vessselinfo file to the disk image, and include an automatic installation rc script. That script would call the Seattle installer, and remove itself once the installer succeeded. After that, the installer will have registered the node manager and software updater to start at boot.
In the meantime, we can make the existing Seattle installation script to be run at boot in such a scenario. It will complain if it finds Seattle already installed, but not hurt otherwise.
The text was updated successfully, but these errors were encountered: