X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/c14246b9a21d038fc6fa850f4032659a98397784..e59b78c774872e33b9c69acc196989a0f565bdf4:/services/keepstore/volume.go diff --git a/services/keepstore/volume.go b/services/keepstore/volume.go index 57e18aba9f..6bce05bec0 100644 --- a/services/keepstore/volume.go +++ b/services/keepstore/volume.go @@ -1,12 +1,33 @@ +// Copyright (C) The Arvados Authors. All rights reserved. +// +// SPDX-License-Identifier: AGPL-3.0 + package main import ( "context" + "crypto/rand" + "fmt" "io" + "math/big" "sync/atomic" "time" + + "git.curoverse.com/arvados.git/sdk/go/arvados" ) +type BlockWriter interface { + // WriteBlock reads all data from r, writes it to a backing + // store as "loc", and returns the number of bytes written. + WriteBlock(ctx context.Context, loc string, r io.Reader) error +} + +type BlockReader interface { + // ReadBlock retrieves data previously stored as "loc" and + // writes it to w. + ReadBlock(ctx context.Context, loc string, w io.Writer) error +} + // A Volume is an interface representing a Keep back-end storage unit: // for example, a single mounted disk, a RAID array, an Amazon S3 volume, // etc. @@ -217,6 +238,13 @@ type Volume interface { // EmptyTrash looks for trashed blocks that exceeded TrashLifetime // and deletes them from the volume. EmptyTrash() + + // Return a globally unique ID of the underlying storage + // device if possible, otherwise "". + DeviceID() string + + // Get the storage classes associated with this volume + GetStorageClasses() []string } // A VolumeWithExamples provides example configs to display in the @@ -229,6 +257,14 @@ type VolumeWithExamples interface { // A VolumeManager tells callers which volumes can read, which volumes // can write, and on which volume the next write should be attempted. type VolumeManager interface { + // Mounts returns all mounts (volume attachments). + Mounts() []*VolumeMount + + // Lookup returns the volume under the given mount + // UUID. Returns nil if the mount does not exist. If + // write==true, returns nil if the volume is not writable. + Lookup(uuid string, write bool) Volume + // AllReadable returns all volumes. AllReadable() []Volume @@ -243,23 +279,71 @@ type VolumeManager interface { // with more free space, etc. NextWritable() Volume + // VolumeStats returns the ioStats used for tracking stats for + // the given Volume. + VolumeStats(Volume) *ioStats + // Close shuts down the volume manager cleanly. Close() } +// A VolumeMount is an attachment of a Volume to a VolumeManager. +type VolumeMount struct { + arvados.KeepMount + volume Volume +} + +// Generate a UUID the way API server would for a "KeepVolumeMount" +// object. +func (*VolumeMount) generateUUID() string { + var max big.Int + _, ok := max.SetString("zzzzzzzzzzzzzzz", 36) + if !ok { + panic("big.Int parse failed") + } + r, err := rand.Int(rand.Reader, &max) + if err != nil { + panic(err) + } + return fmt.Sprintf("zzzzz-ivpuk-%015s", r.Text(36)) +} + // RRVolumeManager is a round-robin VolumeManager: the Nth call to // NextWritable returns the (N % len(writables))th writable Volume // (where writables are all Volumes v where v.Writable()==true). type RRVolumeManager struct { + mounts []*VolumeMount + mountMap map[string]*VolumeMount readables []Volume writables []Volume counter uint32 + iostats map[Volume]*ioStats } // MakeRRVolumeManager initializes RRVolumeManager func MakeRRVolumeManager(volumes []Volume) *RRVolumeManager { - vm := &RRVolumeManager{} + vm := &RRVolumeManager{ + iostats: make(map[Volume]*ioStats), + } + vm.mountMap = make(map[string]*VolumeMount) for _, v := range volumes { + sc := v.GetStorageClasses() + if len(sc) == 0 { + sc = []string{"default"} + } + mnt := &VolumeMount{ + KeepMount: arvados.KeepMount{ + UUID: (*VolumeMount)(nil).generateUUID(), + DeviceID: v.DeviceID(), + ReadOnly: !v.Writable(), + Replication: v.Replication(), + StorageClasses: sc, + }, + volume: v, + } + vm.iostats[v] = &ioStats{} + vm.mounts = append(vm.mounts, mnt) + vm.mountMap[mnt.UUID] = mnt vm.readables = append(vm.readables, v) if v.Writable() { vm.writables = append(vm.writables, v) @@ -268,6 +352,18 @@ func MakeRRVolumeManager(volumes []Volume) *RRVolumeManager { return vm } +func (vm *RRVolumeManager) Mounts() []*VolumeMount { + return vm.mounts +} + +func (vm *RRVolumeManager) Lookup(uuid string, needWrite bool) Volume { + if mnt, ok := vm.mountMap[uuid]; ok && (!needWrite || !mnt.ReadOnly) { + return mnt.volume + } else { + return nil + } +} + // AllReadable returns an array of all readable volumes func (vm *RRVolumeManager) AllReadable() []Volume { return vm.readables @@ -287,18 +383,35 @@ func (vm *RRVolumeManager) NextWritable() Volume { return vm.writables[i%uint32(len(vm.writables))] } +// VolumeStats returns an ioStats for the given volume. +func (vm *RRVolumeManager) VolumeStats(v Volume) *ioStats { + return vm.iostats[v] +} + // Close the RRVolumeManager func (vm *RRVolumeManager) Close() { } -// VolumeStatus provides status information of the volume consisting of: -// * mount_point -// * device_num (an integer identifying the underlying storage system) -// * bytes_free -// * bytes_used +// VolumeStatus describes the current condition of a volume type VolumeStatus struct { - MountPoint string `json:"mount_point"` - DeviceNum uint64 `json:"device_num"` - BytesFree uint64 `json:"bytes_free"` - BytesUsed uint64 `json:"bytes_used"` + MountPoint string + DeviceNum uint64 + BytesFree uint64 + BytesUsed uint64 +} + +// ioStats tracks I/O statistics for a volume or server +type ioStats struct { + Errors uint64 + Ops uint64 + CompareOps uint64 + GetOps uint64 + PutOps uint64 + TouchOps uint64 + InBytes uint64 + OutBytes uint64 +} + +type InternalStatser interface { + InternalStats() interface{} }