storj/satellite/accounting/live/live_test.go

342 lines
8.9 KiB
Go
Raw Normal View History

// Copyright (C) 2019 Storj Labs, Inc.
// See LICENSE for copying information.
package live_test
import (
"context"
"math/rand"
"testing"
"time"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
"go.uber.org/zap/zaptest"
"golang.org/x/sync/errgroup"
"storj.io/common/testcontext"
"storj.io/common/testrand"
"storj.io/common/uuid"
"storj.io/storj/private/testredis"
"storj.io/storj/satellite/accounting"
"storj.io/storj/satellite/accounting/live"
)
func TestAddGetProjectStorageAndBandwidthUsage(t *testing.T) {
tests := []struct {
backend string
}{
{
backend: "redis",
},
}
ctx := testcontext.New(t)
defer ctx.Cleanup()
redis, err := testredis.Start(ctx)
require.NoError(t, err)
defer ctx.Check(redis.Close)
for _, tt := range tests {
tt := tt
t.Run(tt.backend, func(t *testing.T) {
ctx := testcontext.New(t)
var config live.Config
if tt.backend == "redis" {
config = live.Config{
StorageBackend: "redis://" + redis.Addr() + "?db=0",
}
}
cache, err := live.OpenCache(ctx, zaptest.NewLogger(t).Named("live-accounting"), config)
require.NoError(t, err)
defer ctx.Check(cache.Close)
populatedData, err := populateCache(ctx, cache)
require.NoError(t, err)
// make sure all of the "projects" got all space updates and got right totals
for _, pdata := range populatedData {
pdata := pdata
t.Run("storage", func(t *testing.T) {
spaceUsed, err := cache.GetProjectStorageUsage(ctx, pdata.projectID)
require.NoError(t, err)
assert.Equalf(t, pdata.storageSum, spaceUsed, "projectID %v", pdata.projectID)
// upate it again and check
negativeVal := -(rand.Int63n(pdata.storageSum) + 1)
pdata.storageSum += negativeVal
err = cache.AddProjectStorageUsage(ctx, pdata.projectID, negativeVal)
require.NoError(t, err)
spaceUsed, err = cache.GetProjectStorageUsage(ctx, pdata.projectID)
require.NoError(t, err)
assert.EqualValues(t, pdata.storageSum, spaceUsed)
})
t.Run("bandwidth", func(t *testing.T) {
bandwidthUsed, err := cache.GetProjectBandwidthUsage(ctx, pdata.projectID, pdata.bandwidthNow)
require.NoError(t, err)
assert.Equalf(t, pdata.bandwidthSum, bandwidthUsed, "projectID %v", pdata.projectID)
// upate it again and check
negativeVal := -(rand.Int63n(pdata.bandwidthSum) + 1)
pdata.bandwidthSum += negativeVal
err = cache.UpdateProjectBandwidthUsage(ctx, pdata.projectID, negativeVal, time.Second*2, pdata.bandwidthNow)
require.NoError(t, err)
bandwidthUsed, err = cache.GetProjectBandwidthUsage(ctx, pdata.projectID, pdata.bandwidthNow)
require.NoError(t, err)
assert.EqualValues(t, pdata.bandwidthSum, bandwidthUsed)
})
}
})
}
}
func TestGetAllProjectTotals(t *testing.T) {
tests := []struct {
backend string
}{
{
backend: "redis",
},
}
ctx := testcontext.New(t)
defer ctx.Cleanup()
redis, err := testredis.Start(ctx)
require.NoError(t, err)
defer ctx.Check(redis.Close)
for _, tt := range tests {
tt := tt
t.Run(tt.backend, func(t *testing.T) {
ctx := testcontext.New(t)
var config live.Config
if tt.backend == "redis" {
config = live.Config{
StorageBackend: "redis://" + redis.Addr() + "?db=0",
}
}
cache, err := live.OpenCache(ctx, zaptest.NewLogger(t).Named("live-accounting"), config)
satellite/accounting: refactor live accounting to hold current estimated totals live accounting used to be a cache to store writes before they are picked up during the tally iteration, after which the cache is cleared. This created a window in which users could potentially exceed the storage limit. This PR refactors live accounting to hold current estimations of space used per project. This should also reduce DB load since we no longer need to query the satellite DB when checking space used for limiting. The mechanism by which the new live accounting system works is as follows: During the upload of any segment, the size of that segment is added to its respective project total in live accounting. At the beginning of the tally iteration we record the current values in live accounting as `initialLiveTotals`. At the end of the tally iteration we again record the current totals in live accounting as `latestLiveTotals`. The metainfo loop observer in tally allows us to get the project totals from what it observed in metainfo DB which are stored in `tallyProjectTotals`. However, for any particular segment uploaded during the metainfo loop, the observer may or may not have seen it. Thus, we take half of the difference between `latestLiveTotals` and `initialLiveTotals`, and add that to the total that was found during tally and set that as the new live accounting total. Initially, live accounting was storing the total stored amount across all nodes rather than the segment size, which is inconsistent with how we record amounts stored in the project accounting DB, so we have refactored live accounting to record segment size Change-Id: Ie48bfdef453428fcdc180b2d781a69d58fd927fb
2019-10-31 17:27:38 +00:00
require.NoError(t, err)
defer ctx.Check(cache.Close)
projectIDs := make([]uuid.UUID, 1000)
for i := range projectIDs {
projectIDs[i] = testrand.UUID()
err := cache.AddProjectStorageUsage(ctx, projectIDs[i], int64(i))
require.NoError(t, err)
err = cache.UpdateProjectSegmentUsage(ctx, projectIDs[i], int64(i))
require.NoError(t, err)
}
satellite/accounting: refactor live accounting to hold current estimated totals live accounting used to be a cache to store writes before they are picked up during the tally iteration, after which the cache is cleared. This created a window in which users could potentially exceed the storage limit. This PR refactors live accounting to hold current estimations of space used per project. This should also reduce DB load since we no longer need to query the satellite DB when checking space used for limiting. The mechanism by which the new live accounting system works is as follows: During the upload of any segment, the size of that segment is added to its respective project total in live accounting. At the beginning of the tally iteration we record the current values in live accounting as `initialLiveTotals`. At the end of the tally iteration we again record the current totals in live accounting as `latestLiveTotals`. The metainfo loop observer in tally allows us to get the project totals from what it observed in metainfo DB which are stored in `tallyProjectTotals`. However, for any particular segment uploaded during the metainfo loop, the observer may or may not have seen it. Thus, we take half of the difference between `latestLiveTotals` and `initialLiveTotals`, and add that to the total that was found during tally and set that as the new live accounting total. Initially, live accounting was storing the total stored amount across all nodes rather than the segment size, which is inconsistent with how we record amounts stored in the project accounting DB, so we have refactored live accounting to record segment size Change-Id: Ie48bfdef453428fcdc180b2d781a69d58fd927fb
2019-10-31 17:27:38 +00:00
usage, err := cache.GetAllProjectTotals(ctx)
require.NoError(t, err)
require.Len(t, usage, len(projectIDs))
// make sure each project ID and total was received
for _, projID := range projectIDs {
totalStorage, err := cache.GetProjectStorageUsage(ctx, projID)
require.NoError(t, err)
assert.Equal(t, totalStorage, usage[projID].Storage)
totalSegments, err := cache.GetProjectSegmentUsage(ctx, projID)
require.NoError(t, err)
assert.Equal(t, totalSegments, usage[projID].Segments)
}
})
}
}
func TestLiveAccountingCache_ProjectBandwidthUsage_expiration(t *testing.T) {
tests := []struct {
backend string
}{
{
backend: "redis",
},
}
ctx := testcontext.New(t)
defer ctx.Cleanup()
redis, err := testredis.Start(ctx)
require.NoError(t, err)
defer ctx.Check(redis.Close)
for _, tt := range tests {
tt := tt
t.Run(tt.backend, func(t *testing.T) {
ctx := testcontext.New(t)
var config live.Config
if tt.backend == "redis" {
config = live.Config{
StorageBackend: "redis://" + redis.Addr() + "?db=0",
}
}
cache, err := live.OpenCache(ctx, zaptest.NewLogger(t).Named("live-accounting"), config)
require.NoError(t, err)
defer ctx.Check(cache.Close)
var (
projectID = testrand.UUID()
now = time.Now()
)
err = cache.UpdateProjectBandwidthUsage(ctx, projectID, rand.Int63n(4096)+1, time.Second, now)
require.NoError(t, err)
if tt.backend == "redis" {
redis.FastForward(time.Second)
}
time.Sleep(2 * time.Second)
_, err = cache.GetProjectBandwidthUsage(ctx, projectID, now)
require.Error(t, err)
})
}
}
func TestInsertProjectBandwidthUsage(t *testing.T) {
tests := []struct {
backend string
}{
{
backend: "redis",
},
}
ctx := testcontext.New(t)
defer ctx.Cleanup()
redis, err := testredis.Start(ctx)
require.NoError(t, err)
defer ctx.Check(redis.Close)
for _, tt := range tests {
tt := tt
t.Run(tt.backend, func(t *testing.T) {
ctx := testcontext.New(t)
var config live.Config
if tt.backend == "redis" {
config = live.Config{
StorageBackend: "redis://" + redis.Addr() + "?db=0",
}
}
cache, err := live.OpenCache(ctx, zaptest.NewLogger(t).Named("live-accounting"), config)
require.NoError(t, err)
defer ctx.Check(cache.Close)
var (
projectID = testrand.UUID()
now = time.Now()
)
expVal := rand.Int63n(4096) + 1
inserted, err := cache.InsertProjectBandwidthUsage(ctx, projectID, expVal, time.Second, now)
require.NoError(t, err)
assert.True(t, inserted, "is inserted")
val, err := cache.GetProjectBandwidthUsage(ctx, projectID, now)
require.NoError(t, err)
require.Equal(t, expVal, val, "inserted value")
{ // This shouldn't set the value because it already exists.
newVal := rand.Int63n(4096) + 1
inserted, err := cache.InsertProjectBandwidthUsage(ctx, projectID, newVal, time.Second, now)
require.NoError(t, err)
assert.False(t, inserted, "is inserted")
val, err := cache.GetProjectBandwidthUsage(ctx, projectID, now)
require.NoError(t, err)
require.Equal(t, expVal, val, "inserted value")
}
if tt.backend == "redis" {
redis.FastForward(time.Second)
}
time.Sleep(2 * time.Second)
_, err = cache.GetProjectBandwidthUsage(ctx, projectID, now)
require.Error(t, err)
})
}
}
type populateCacheData struct {
projectID uuid.UUID
storageSum int64
bandwidthSum int64
bandwidthNow time.Time
}
func populateCache(ctx context.Context, cache accounting.Cache) ([]populateCacheData, error) {
var (
valuesListSize = rand.Intn(10) + 10
numProjects = rand.Intn(100) + 100
valueStorageMultiplier = rand.Int63n(4095) + 1
valueBandwdithMultiplier = rand.Int63n(4095) + 1
)
// make a largish list of varying values
baseValues := make([]int64, valuesListSize)
for i := range baseValues {
baseValues[i] = rand.Int63n(int64(valuesListSize)) + 1
}
// make up some project IDs
populatedData := make([]populateCacheData, numProjects)
for i := range populatedData {
populatedData[i] = populateCacheData{
projectID: testrand.UUID(),
}
}
// send lots of space used updates for all of these projects to the live
// accounting store.
errg, ctx := errgroup.WithContext(context.Background())
for i, pdata := range populatedData {
var (
i = i
projID = pdata.projectID
)
errg.Go(func() error {
// have each project sending the values in a different order
myValues := make([]int64, valuesListSize)
copy(myValues, baseValues)
rand.Shuffle(valuesListSize, func(v1, v2 int) {
myValues[v1], myValues[v2] = myValues[v2], myValues[v1]
})
now := time.Now()
populatedData[i].bandwidthNow = now
for _, val := range myValues {
storageVal := val * valueStorageMultiplier
populatedData[i].storageSum += storageVal
if err := cache.AddProjectStorageUsage(ctx, projID, storageVal); err != nil {
return err
}
bandwidthVal := val * valueBandwdithMultiplier
populatedData[i].bandwidthSum += bandwidthVal
if err := cache.UpdateProjectBandwidthUsage(ctx, projID, bandwidthVal, time.Hour, now); err != nil {
return err
}
}
return nil
})
}
return populatedData, errg.Wait()
satellite/accounting: refactor live accounting to hold current estimated totals live accounting used to be a cache to store writes before they are picked up during the tally iteration, after which the cache is cleared. This created a window in which users could potentially exceed the storage limit. This PR refactors live accounting to hold current estimations of space used per project. This should also reduce DB load since we no longer need to query the satellite DB when checking space used for limiting. The mechanism by which the new live accounting system works is as follows: During the upload of any segment, the size of that segment is added to its respective project total in live accounting. At the beginning of the tally iteration we record the current values in live accounting as `initialLiveTotals`. At the end of the tally iteration we again record the current totals in live accounting as `latestLiveTotals`. The metainfo loop observer in tally allows us to get the project totals from what it observed in metainfo DB which are stored in `tallyProjectTotals`. However, for any particular segment uploaded during the metainfo loop, the observer may or may not have seen it. Thus, we take half of the difference between `latestLiveTotals` and `initialLiveTotals`, and add that to the total that was found during tally and set that as the new live accounting total. Initially, live accounting was storing the total stored amount across all nodes rather than the segment size, which is inconsistent with how we record amounts stored in the project accounting DB, so we have refactored live accounting to record segment size Change-Id: Ie48bfdef453428fcdc180b2d781a69d58fd927fb
2019-10-31 17:27:38 +00:00
}