D
Deleted member 205422
Guest
So I am completely aware this would be considered officially unsupported as already installing PVE on top of Debian is, however I basically would like to hear from someone who attempted the same or abandoned doing so and why.
The logic behind iPXE booting multiple nodes is ease of management - no need for OOB features on commodity hardware, some nodes being possibly completely diskless, others only use fast NVMe entirely as ZFS pools for VMs/CTs.
It appears root filesystem of PVE alone is well under 2GB, this is completely doable from ramdisk, so the whole filesystem could be copied over under 20 seconds onto ramdisk and off it goes. Another benefit would be that e.g. OS update should be possible once for all nodes, next time they reboot they are all up to date. Logging goes out of the syslog, telemetry into InfluxDB. What am I missing and why isn't this supported for large scale deployments?
MOST IMPORTANTLY: Which paths are worth keeping e.g. periodically saved back onto the NFS volume as they are config-specific to each individual node?
Thanks everyone.
The logic behind iPXE booting multiple nodes is ease of management - no need for OOB features on commodity hardware, some nodes being possibly completely diskless, others only use fast NVMe entirely as ZFS pools for VMs/CTs.
It appears root filesystem of PVE alone is well under 2GB, this is completely doable from ramdisk, so the whole filesystem could be copied over under 20 seconds onto ramdisk and off it goes. Another benefit would be that e.g. OS update should be possible once for all nodes, next time they reboot they are all up to date. Logging goes out of the syslog, telemetry into InfluxDB. What am I missing and why isn't this supported for large scale deployments?
MOST IMPORTANTLY: Which paths are worth keeping e.g. periodically saved back onto the NFS volume as they are config-specific to each individual node?
Thanks everyone.
