frostfs-sdk-go/netmap/netmap.go

301 lines
7.3 KiB
Go

package netmap
import (
"fmt"
"git.frostfs.info/TrueCloudLab/frostfs-api-go/v2/netmap"
"git.frostfs.info/TrueCloudLab/hrw"
)
// NetMap represents FrostFS network map. It includes information about all
// storage nodes registered in FrostFS the network.
//
// NetMap is mutually compatible with git.frostfs.info/TrueCloudLab/frostfs-api-go/v2/netmap.NetMap
// message. See ReadFromV2 / WriteToV2 methods.
//
// Instances can be created using built-in var declaration.
type NetMap struct {
epoch uint64
nodes []NodeInfo
}
// ReadFromV2 reads NetMap from the netmap.NetMap message. Checks if the
// message conforms to FrostFS API V2 protocol.
//
// See also WriteToV2.
func (m *NetMap) ReadFromV2(msg netmap.NetMap) error {
var err error
nodes := msg.Nodes()
if nodes == nil {
m.nodes = nil
} else {
m.nodes = make([]NodeInfo, len(nodes))
for i := range nodes {
err = m.nodes[i].ReadFromV2(nodes[i])
if err != nil {
return fmt.Errorf("invalid node info: %w", err)
}
}
}
m.epoch = msg.Epoch()
return nil
}
// WriteToV2 writes NetMap to the netmap.NetMap message. The message
// MUST NOT be nil.
//
// See also ReadFromV2.
func (m NetMap) WriteToV2(msg *netmap.NetMap) {
var nodes []netmap.NodeInfo
if m.nodes != nil {
nodes = make([]netmap.NodeInfo, len(m.nodes))
for i := range m.nodes {
m.nodes[i].WriteToV2(&nodes[i])
}
msg.SetNodes(nodes)
}
msg.SetEpoch(m.epoch)
}
// SetNodes sets information list about all storage nodes from the FrostFS network.
//
// Argument MUST NOT be mutated, make a copy first.
//
// See also Nodes.
func (m *NetMap) SetNodes(nodes []NodeInfo) {
m.nodes = nodes
}
// Nodes returns nodes set using SetNodes.
//
// Return value MUST not be mutated, make a copy first.
func (m NetMap) Nodes() []NodeInfo {
return m.nodes
}
// SetEpoch specifies revision number of the NetMap.
//
// See also Epoch.
func (m *NetMap) SetEpoch(epoch uint64) {
m.epoch = epoch
}
// Epoch returns epoch set using SetEpoch.
//
// Zero NetMap has zero revision.
func (m NetMap) Epoch() uint64 {
return m.epoch
}
// nodes is a slice of NodeInfo instances needed for HRW sorting.
type nodes []NodeInfo
// assert nodes type provides hrw.Hasher required for HRW sorting.
var _ hrw.Hasher = nodes{}
// Hash is a function from hrw.Hasher interface. It is implemented
// to support weighted hrw sorting of buckets. Each bucket is already sorted by hrw,
// thus giving us needed "randomness".
func (n nodes) Hash() uint64 {
if len(n) > 0 {
return n[0].Hash()
}
return 0
}
// weights returns slice of nodes weights W.
func (n nodes) weights(wf weightFunc) []float64 {
w := make([]float64, 0, len(n))
for i := range n {
w = append(w, wf(n[i]))
}
return w
}
func flattenNodes(ns []nodes) nodes {
var sz, i int
for i = range ns {
sz += len(ns[i])
}
result := make(nodes, 0, sz)
for i := range ns {
result = append(result, ns[i]...)
}
return result
}
// PlacementVectors sorts container nodes returned by ContainerNodes method
// and returns placement vectors for the entity identified by the given pivot.
// For example, in order to build node list to store the object, binary-encoded
// object identifier can be used as pivot. Result is deterministic for
// the fixed NetMap and parameters.
func (m NetMap) PlacementVectors(vectors [][]NodeInfo, pivot []byte) ([][]NodeInfo, error) {
h := hrw.Hash(pivot)
wf := defaultWeightFunc(m.nodes)
result := make([][]NodeInfo, len(vectors))
for i := range vectors {
result[i] = make([]NodeInfo, len(vectors[i]))
copy(result[i], vectors[i])
hrw.SortHasherSliceByWeightValue(result[i], nodes(result[i]).weights(wf), h)
}
return result, nil
}
// SelectFilterNodes returns a two-dimensional list of nodes as a result of applying the
// given SelectFilterExpr to the NetMap.
// If the SelectFilterExpr contains only filters, the result contains a single row with the
// result of the last filter application.
// If the SelectFilterExpr contains only selectors, the result contains the selection rows
// of the last select application.
func (m NetMap) SelectFilterNodes(expr *SelectFilterExpr) ([][]NodeInfo, error) {
p := PlacementPolicy{
filters: expr.filters,
}
if expr.selector != nil {
p.selectors = append(p.selectors, *expr.selector)
}
c := newContext(m)
c.setCBF(expr.cbf)
if err := c.processFilters(p); err != nil {
return nil, err
}
if err := c.processSelectors(p); err != nil {
return nil, err
}
if expr.selector == nil {
var ret []NodeInfo
lastFilter := expr.filters[len(expr.filters)-1]
for _, ni := range m.nodes {
if c.match(c.processedFilters[lastFilter.GetName()], ni) {
ret = append(ret, ni)
}
}
return [][]NodeInfo{ret}, nil
}
sel, err := c.getSelection(*c.processedSelectors[expr.selector.GetName()])
if err != nil {
return nil, err
}
var ret [][]NodeInfo
for i, ns := range sel {
ret = append(ret, []NodeInfo{})
for _, n := range ns {
ret[i] = append(ret[i], n)
}
}
return ret, nil
}
func countNodes(r netmap.Replica) uint32 {
if r.GetCount() != 0 {
return r.GetCount()
}
return r.GetECDataCount() + r.GetECParityCount()
}
func (p PlacementPolicy) isUnique() bool {
if p.unique {
return true
}
for _, r := range p.replicas {
if r.GetECDataCount() != 0 || r.GetECParityCount() != 0 {
return true
}
}
return false
}
// ContainerNodes returns two-dimensional list of nodes as a result of applying
// given PlacementPolicy to the NetMap. Each line of the list corresponds to a
// replica descriptor. Line order corresponds to order of ReplicaDescriptor list
// in the policy. Nodes are pre-filtered according to the Filter list from
// the policy, and then selected by Selector list. Result is deterministic for
// the fixed NetMap and parameters.
//
// Result can be used in PlacementVectors.
func (m NetMap) ContainerNodes(p PlacementPolicy, pivot []byte) ([][]NodeInfo, error) {
c := newContext(m)
c.setPivot(pivot)
c.setCBF(p.backupFactor)
if err := c.processFilters(p); err != nil {
return nil, err
}
if err := c.processSelectors(p); err != nil {
return nil, err
}
unique := p.isUnique()
result := make([][]NodeInfo, len(p.replicas))
// Note that the cached selectors are not used when the policy contains the UNIQUE flag.
// This is necessary because each selection vector affects potentially the subsequent vectors
// and thus we call getSelection in such case, in order to take into account nodes previously
// marked as used by earlier replicas.
for i := range p.replicas {
sName := p.replicas[i].GetSelector()
if sName == "" && !(len(p.replicas) == 1 && len(p.selectors) == 1) {
var s netmap.Selector
s.SetCount(countNodes(p.replicas[i]))
s.SetFilter(mainFilterName)
nodes, err := c.getSelection(s)
if err != nil {
return nil, err
}
result[i] = append(result[i], flattenNodes(nodes)...)
if unique {
c.addUsedNodes(result[i]...)
}
continue
}
if unique {
if c.processedSelectors[sName] == nil {
return nil, fmt.Errorf("selector not found: '%s'", sName)
}
nodes, err := c.getSelection(*c.processedSelectors[sName])
if err != nil {
return nil, err
}
result[i] = append(result[i], flattenNodes(nodes)...)
c.addUsedNodes(result[i]...)
} else {
nodes, ok := c.selections[sName]
if !ok {
return nil, fmt.Errorf("selector not found: REPLICA '%s'", sName)
}
result[i] = append(result[i], flattenNodes(nodes)...)
}
}
return result, nil
}