Storage on Controller Hosts¶
The controller’s root disk provides storage for the StarlingX system databases, system configuration files, local docker images, container’s ephemeral filesystems, the Local Docker Registry container image store, platform backup, and the system backup operations.
Container local storage is derived from the cgts-vg volume group on the root disk. You can add additional storage to the cgts-vg volume by assigning a partition or disk to make it larger. This will allow you to increase the size of the container local storage for the host, however, you cannot assign it specifically to a non-root disk.
On All-in-one Simplex, All-in-one Duplex, and Standard with controller storage systems, at least one additional disk for each controller host is required for backing container PVCs.
Two disks are required with one being a Ceph OSD.
Root Filesystem Storage¶
Space on the root disk is allocated to provide filesystem storage.
You can increase the allotments for the following filesystems using the Horizon Web interface or the CLI. The following commands are available to increase various filesystem sizes; system controllerfs, and system host-fs.
Synchronized filesystems ensure that files stored in several different physical locations are up to date. The following commands can be used to resize an DRBD-synced filesystem (Database, Docker-distribution, Etcd, Extension, Platform) on controllers: controllerfs-list, controllerfs-modify, and controllerfs-show.
This is the storage allotment for a variety of platform items including the local helm repository, the StarlingX application repository, and internal platform configuration data files.
The storage allotment for the platform’s postgres database is used by StarlingX, System Inventory, Keystone and Barbican.
Internal database storage is provided using DRBD-synchronized partitions on the controller primary disks. The size of the database grows with the number of system resources created by the system administrator. This includes objects of all kinds such as hosts, interfaces, and service parameters.
If you add a database filesystem or increase its size, you must also increase the size of the backup filesystem.
Docker-distribution Storage (Local Docker Registry storage)
The storage allotment for container images stored in the local docker registry. This storage is provided using a DRBD-synchronized partition on the controller primary disk.
The storage allotment for the Kubernetes etcd database.
Internal database storage is provided using a DRBD-synchronized partition on the controller primary disk. The size of the database grows with the number of system resources created by the system administrator and the users. This includes objects of all kinds such as pods, services, and secrets.
Ceph-mon is the cluster monitor daemon for the Ceph distributed file system that is used for Ceph monitors to synchronize.
This filesystem is reserved for future use. This storage is implemented on a DRBD-synchronized partition on the controller primary disk.
The following host filesystem commands can be used to resize non-DRBD filesystems (Backup, Docker, Kubelet and Scratch) and do not apply to all hosts of a give personality type:
host-fs-list, host-fs-modify, and host-fs-show
The host-fs-modify command increases the storage configuration for the filesystem specified on a per-host basis. For example, the following command increases the scratch filesystem size to 10 GB:
~(keystone_admin)]$ system host-fs-modify controller-1 scratch=10
This is the storage allotment for backup operations. This is a backup area, where, backup=2*database+platform size.
This storage allotment is for ephemeral filesystems for containers on the host, and for docker image cache.
This storage allotment is for ephemeral storage size related to kubernetes pods on this host.
This storage allotment is used by the host as a temp area for a variety of miscellaneous transient host operations.
This is the storage allotment for log data. This filesystem is not resizable. Logs are rotated within the fixed space allocated.
Replacement root disks for a reinstalled controller should be the same size or larger to ensure that existing allocation sizes for filesystems will fit on the replacement disk.
Persistent Volume Claims Storage (Ceph Cluster)¶
For controller-storage systems, additional disks on the controller, configured as Ceph OSDs, provide a small Ceph cluster for backing Persistent Volume Claims storage for Containers.
On AIO-SX systems, replication is done between OSDs within the host.
The following three replication factors are supported:
This is the default, and requires one or more OSD disks.
This requires two or more OSD disks.
This requires three or more OSD disks.
On AIO-DX systems, replication is between the two controllers. Only one replication group is supported and additional controllers cannot be added.
The following replication factor is supported:
There can be any number of OSDs on each controller, with a minimum of one each. It is recommended that you use the same number and same size OSD disks on the controllers.