Documentation
¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func CheckStoragePlansAlike ¶ added in v0.30.0
func CheckStoragePlansAlike(storagePlans []*StoragePlan) error
By alikeness, I mean, the disks across which the ZFS pool will be running, must be of the same count and type (SSD / HDD / NVMe) across all the nodes in the node-group.
All the nodes in a node-group share a single KubeadmControlPlane / KubeadmConfig resource, whose preKubeadm section runs the same `kubeaid-storagectl plan execute --os-size ... --zfs-pool-size ...` command on every node. Each node recomputes its own storage plan locally, so device names (sdc vs sdd) are free to differ between nodes — the kernel assigns them from port enumeration order, which isn't stable across otherwise identical machines. What must match is the resulting pool's shape: a mirror across 2 SSDs on one node and across 2 HDDs on another would give the node-group wildly different storage performance while being labelled identically.
Types ¶
type Disk ¶
type Disk struct {
Name,
WWN,
Type,
PartitionTableType string
Size int // GB.
// Whether the server to which this disk is attached, has a NIC (Network Interface Card)
// with speed >= 5 GBPS.
// If yes, and this disk is an SSD / NVMe, then the priority score for CEPH installation
// increases : since the CEPH OSDs can leverage the increased network bandwidth for higher
// disk IOPS.
WithHighSpeedNIC bool
PriorityScores PriorityScores
Allocations struct {
OS,
ZFS,
CEPH int // GB.
}
}
func NewDisk ¶
NewDisk constructs a Disk with all identification + sizing fields set and allocations / priority scores zero-initialised.
func (*Disk) AssignPriorityScores ¶
func (d *Disk) AssignPriorityScores()
Assigns priority scores to the disk, for OS and ZFS installations.
func (*Disk) Unallocated ¶
Returns the amount of unallocated storage.
type NodeGroupName ¶
type NodeGroupName = string
type PriorityScores ¶
type PriorityScores struct {
OS,
ZFS int
}
type StoragePlan ¶
func (*StoragePlan) Execute ¶
func (s *StoragePlan) Execute(ctx context.Context, commandExecutor commandexecutor.CommandExecutor)
Executes the storage plan, by running necessary shell commands.
func (*StoragePlan) PrettyPrint ¶
func (s *StoragePlan) PrettyPrint()
type StoragePlanExecutorTemplateValues ¶
type StoragePlanExecutorTemplateValues struct {
StoragePlan *StoragePlan
}
type StoragePlans ¶
type StoragePlans map[NodeGroupName][]*StoragePlan
func (*StoragePlans) GetApproval ¶
func (s *StoragePlans) GetApproval(ctx context.Context)
func (StoragePlans) PrettyPrint ¶
func (s StoragePlans) PrettyPrint()
PrettyPrint renders the storage layout the operator is about to approve as a single bordered box — one section per node group, each with an underlined uppercase header and a right-aligned size column — followed by a destructive-change warning.
AreStoragePlansAlike enforces an identical layout across every server in a group, so the first plan is canonical and the box shows per-group facts, not per-server ones. Groups render in a stable order (control-plane first, then alphabetical) so re-running the command produces the same screen.
Source Files