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
By default installing an /etc/hostid file in the post install script has proven to be a problem for diskless systems which share a common image. This results in all the diskless nodes for a cluster having the same hostid which causes failures with MPI applications. To fix this I suggest that we not automatically create an /etc/hostid on the system itself, Instead extend the dracut module-setup.sh script to dynamically create an /etc/hostid only for the initramfs which matches the default hostid for the system. This should still allow diskless booting and avoid breakage in diskless environments.
The text was updated successfully, but these errors were encountered:
What about a mac->hostid lookup file in the base image. look up the mac of our eth interfaces, get out the first corresponding hostid we find. Smart admins know to only add one MAC per machine
By default installing an /etc/hostid file in the post install script has proven to be a problem for diskless systems which share a common image. This results in all the diskless nodes for a cluster having the same hostid which causes failures with MPI applications. To fix this I suggest that we not automatically create an /etc/hostid on the system itself, Instead extend the dracut module-setup.sh script to dynamically create an /etc/hostid only for the initramfs which matches the default hostid for the system. This should still allow diskless booting and avoid breakage in diskless environments.
The text was updated successfully, but these errors were encountered: