This message was deleted.
# harvester
This message was deleted.
depends on the use case.. how many nodes you will have in the cluster.. and in turn if you will need a non default storage class with different replica counts
may be good to have mirroring on a smaller disk of say 200G which is used for the installs and contains the OS info, and use a striped disk for additional data disk, assuming you will let longhorn do the replication
then the striped disk could be used for additional data disks..
Thanks for sharing your knowledge! I am thinking at least a 6 node cluster, probably will add more as we go. I think we will be using 1-2TB drives and maybe even 4TB drives to ensure enough room for databases. The Dell servers we have use a RAID in the BIOS and not sure how to check to see if a drive failed without shutting down Harvester and going into the BIOS
@bulky-lion-74983 did you managed to install Harvester OS on a Software RAID? (I’m searching a solution for this, too -
@blue-kitchen-51801 I have not done that yet. We did try a hardware raid card for all the drives. I am guessing you are looking to protect the Harvester OS from single disk failure taking down the entire machine. One of the things we were looking at was using the hardware raid and then using the PCI passthrough to get access to the hardware raid card from a VM, allowing a utility to detect a disk failure so we can fix it without taking down the machine to the bios. In order for software raid to work it would need to be part of Harvester. Interesting idea!
normally, it shouldn’t be very complicated to implement it, but I don’t have the context and the expertise to implement it • here we will need to remove
grep -w disk
- • then, we will need to not tear down LVM and MD (maybe display a warning or something like this) - • from what I’ve tested
elemental install
partition, not sure why.. -[…]8 I hope that someone from Rancher picks this up.