4fb8adb9e4
* Makes local volume provisioner more dynamic * Correct variable name in local storage provisioner defaults * Updates external-provisioner readme * Updates variable naming to be more clear, more documentation, fixes sample inventory * Variable refactor, untangled some jinja2 loops * Corrects variable name * No variable substitution in dict keys, replaced with anchor * Fixes default storage_classes dict, inline docs * Fixes spelling in inline docs * Addresses comments in review * Updates all the defaults * Fix failing CI task * Fixes external provisioner daemonset |
||
---|---|---|
.. | ||
defaults | ||
tasks | ||
templates | ||
README.md |
Local Storage Provisioner
The local storage provisioner
is NOT a dynamic storage provisioner as you would
expect from a cloud provider. Instead, it simply creates PersistentVolumes for
all mounts under the host_dir of the specified storage class.
These storage classes are specified in the local_volume_provisioner_storage_classes
list.
An example this list:
local_volume_provisioner_storage_classes:
- local-storage:
host_dir: /mnt/disks
mount_dir: /mnt/disks
- fast-disks:
host_dir: /mnt/fast-disks
mount_dir: /mnt/fast-disks
block_cleaner_command:
- "/scripts/shred.sh"
- "2"
volume_mode: Filesystem
fs_type: ext4
For each dictionary in local_volume_provisioner_storage_classes
a storageClass with the
same name is created. The keys of this dictionary are converted to camelCase and added
as attributes to the storageClass.
The result of the above example is:
data:
storageClassMap: |
local-storage:
hostDir: /mnt/disks
mountDir: /mnt/disks
fast-disks:
hostDir: /mnt/fast-disks
mountDir: /mnt/fast-disks
blockCleanerCommand:
- "/scripts/shred.sh"
- "2"
volumeMode: Filesystem
fsType: ext4
The default StorageClass is local-storage on /mnt/disks, the rest of this doc will use that path as an example.
Examples to create local storage volumes
tmpfs method:
for vol in vol1 vol2 vol3; do
mkdir /mnt/disks/$vol
mount -t tmpfs -o size=5G $vol /mnt/disks/$vol
done
The tmpfs method is not recommended for production because the mount is not persistent and data will be deleted on reboot.
Mount physical disks
mkdir /mnt/disks/ssd1
mount /dev/vdb1 /mnt/disks/ssd1
Physical disks are recommended for production environments because it offers complete isolation in terms of I/O and capacity.
File-backed sparsefile method
truncate /mnt/disks/disk5 --size 2G
mkfs.ext4 /mnt/disks/disk5
mkdir /mnt/disks/vol5
mount /mnt/disks/disk5 /mnt/disks/vol5
If you have a development environment and only one disk, this is the best way to limit the quota of persistent volumes.
Simple directories
In a development environment using mount --bind
works also, but there is no capacity
management.
Block volumeMode PVs
Create a symbolic link under discovery directory to the block device on the node. To use raw block devices in pods BlockVolume feature gate must be enabled.
Usage notes
Beta PV.NodeAffinity field is used by default. If running against an older K8s version, the useAlphaAPI flag must be set in the configMap.
The volume provisioner cannot calculate volume sizes correctly, so you should delete the daemonset pod on the relevant host after creating volumes. The pod will be recreated and read the size correctly.
Make sure to make any mounts persist via /etc/fstab or with systemd mounts (for CoreOS/Container Linux). Pods with persistent volume claims will not be able to start if the mounts become unavailable.
Further reading
Refer to the upstream docs here: https://github.com/kubernetes-incubator/external-storage/tree/master/local-volume