Rook/Ceph on ROS Bare Metal is missing drives

RancherOS
1

I try to deploy rook/ceph chart via catalog on Bare Metal and OS is ROS. Here is the link for instruction. https://multinode-kubernetes-cluster.readthedocs.io/en/latest/03-k8s-helm_and_packages.html

The rook/ceph cluster can get /mnt/sda only if I follow up the instruction as above but not other drives /mnt/sdb, /mnt/sdc, /mnt/sdd, /mnt/sde and /mnt/sdf.

I try to turn on useAllDevices: True but it will even missing /mnt/sda which is no osd at all.
e.g.

  storage: # cluster level storage configuration and selection
    useAllNodes: true
    useAllDevices: true

Does anyone have experience how to handle it ?

Here has some drive info fyi.

$ blkid
/dev/sda1: LABEL="RANCHER_STATE" UUID="29785853-9b57-4268-be69-9d6f580c8235" TYPE="ext4" PARTUUID="33be3f85-01"
/dev/sdb: UUID="d7cc10a2-9a60-4da2-b535-4584040fba63" TYPE="ext4"
/dev/sdc: UUID="6e1323fb-e26a-4f73-85dd-beb2d3fa7081" TYPE="ext4"
/dev/sdd: UUID="c1494bdf-e186-44b0-b465-0fd4d2bccf16" TYPE="ext4"
/dev/sde: UUID="9d269f7a-be30-4275-8d4a-fe57b39509a4" TYPE="ext4"
/dev/sdf: UUID="c115172c-f492-4f90-9613-d30b8cf6ca3b" TYPE="ext4"
/dev/sdg: UUID="WlQqjg-UgJk-rxPQ-ZSuZ-sWIe-f09K-Yat3DT" TYPE="LVM2_member"

Iā€™m wondering does ROS handle drives in different way and need some extra steps to make it visitable from k8s ?
e.g. ā€œhttps://rancher.com/docs/os/v1.2/en/storage/state-partition/ā€

2

@chianingwang Try system-docker restart udev-cold and system-docker restart udev

1 Like