Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[coord] Guard against duplicate __rollup__ tags #3950

Merged
merged 4 commits into from
Dec 1, 2021
Merged
Show file tree
Hide file tree
Changes from 1 commit
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
172 changes: 77 additions & 95 deletions src/cmd/services/m3coordinator/downsample/id_pool_types.go
Original file line number Diff line number Diff line change
Expand Up @@ -65,25 +65,21 @@ func isRollupID(
// method, it will return the rollup tag in the correct alphabetical order
// when progressing through the existing tags.
type rollupIDProvider struct {
index int
newName []byte
tagPairs []id.TagPair
nameTagIndex int
rollupTagIndex int

tagEncoder serialize.TagEncoder
pool *rollupIDProviderPool
nameTag ident.ID
nameTagBytes []byte
nameTagBeforeRollupTag bool
tagNameID *ident.ReusableBytesID
tagValueID *ident.ReusableBytesID
mergeTags []mergeTag
}

type mergeTag struct {
name []byte
index int
index int
len int
mergeTagsIdx int
newName []byte
tagPairs []id.TagPair
curr id.TagPair
currIdx int

tagEncoder serialize.TagEncoder
pool *rollupIDProviderPool
nameTag ident.ID
nameTagBytes []byte
tagNameID *ident.ReusableBytesID
tagValueID *ident.ReusableBytesID
mergeTags []id.TagPair
}

func newRollupIDProvider(
Expand All @@ -93,29 +89,27 @@ func newRollupIDProvider(
) *rollupIDProvider {
nameTagBytes := nameTag.Bytes()
nameTagBeforeRollupTag := bytes.Compare(nameTagBytes, rollupTagName) < 0
mergeTags := []mergeTag{
mergeTags := []id.TagPair{
{
name: rollupTagName,
index: -1,
Name: rollupTagName,
Value: rollupTagValue,
},
{
name: nameTagBytes,
index: -1,
Name: nameTagBytes,
// Value is set in reset
},
}
if nameTagBeforeRollupTag {
mergeTags[0].name = nameTagBytes
mergeTags[1].name = rollupTagName
mergeTags[0], mergeTags[1] = mergeTags[1], mergeTags[0]
}
return &rollupIDProvider{
tagEncoder: tagEncoder,
pool: pool,
nameTag: nameTag,
nameTagBytes: nameTagBytes,
nameTagBeforeRollupTag: nameTagBeforeRollupTag,
tagNameID: ident.NewReusableBytesID(),
tagValueID: ident.NewReusableBytesID(),
mergeTags: mergeTags,
tagEncoder: tagEncoder,
pool: pool,
nameTag: nameTag,
nameTagBytes: nameTagBytes,
tagNameID: ident.NewReusableBytesID(),
tagValueID: ident.NewReusableBytesID(),
mergeTags: mergeTags,
}
}

Expand Down Expand Up @@ -143,41 +137,26 @@ func (p *rollupIDProvider) reset(
newName []byte,
tagPairs []id.TagPair,
) {
p.index = -1
p.index = 0
p.mergeTagsIdx = 0
p.newName = newName
p.tagPairs = tagPairs
p.mergeTags[0].index = -1
p.mergeTags[1].index = -1

// merge the special tags into the set of tag pairs without allocating extra space.
idx := 0
for _, pair := range tagPairs {
if p.mergeTags[0].index == -1 && bytes.Compare(p.mergeTags[0].name, pair.Name) < 0 {
p.mergeTags[0].index = idx
idx++
p.currIdx = -1

var dups int
// precompute the length of the "combined" slice for the Len() method.
// mergeTags is small so it's fine to do n^2 instead of a more complicated O(n) merge scan.
for i := range p.tagPairs {
for j := range p.mergeTags {
if bytes.Compare(p.tagPairs[i].Name, p.mergeTags[j].Name) == 0 {
dups++
}
if bytes.Compare(p.mergeTags[j].Name, p.nameTagBytes) == 0 {
p.mergeTags[j].Value = newName
}
}
if p.mergeTags[1].index == -1 && bytes.Compare(p.mergeTags[1].name, pair.Name) < 0 {
p.mergeTags[1].index = idx
// all tags merged, safe to break.
break
}
idx++
}

if p.mergeTags[0].index == -1 {
p.mergeTags[0].index = idx
idx++
}
if p.mergeTags[1].index == -1 {
p.mergeTags[1].index = idx
}

p.rollupTagIndex = p.mergeTags[0].index
p.nameTagIndex = p.mergeTags[1].index
if p.nameTagBeforeRollupTag {
p.nameTagIndex = p.mergeTags[0].index
p.rollupTagIndex = p.mergeTags[1].index
}
p.len = len(p.tagPairs) + len(p.mergeTags) - dups
}

func (p *rollupIDProvider) finalize() {
Expand All @@ -186,42 +165,45 @@ func (p *rollupIDProvider) finalize() {
}
}

// Next takes the smallest element across both sets of tags, removing any duplicates between the lists.
func (p *rollupIDProvider) Next() bool {
p.index++
return p.index < p.Len()
if p.index == len(p.tagPairs) && p.mergeTagsIdx == len(p.mergeTags) {
// at the end of both sets
return false
}
if p.index == len(p.tagPairs) {
// only merged tags left
p.curr = p.mergeTags[p.mergeTagsIdx]
p.mergeTagsIdx++
} else if p.mergeTagsIdx == len(p.mergeTags) {
// only provided tags left
p.curr = p.tagPairs[p.index]
p.index++
} else if bytes.Compare(p.tagPairs[p.index].Name, p.mergeTags[p.mergeTagsIdx].Name) == 0 {
// a merge tag exists in the provided tag, advance both to prevent duplicates.
p.curr = p.tagPairs[p.index]
p.index++
p.mergeTagsIdx++
} else if bytes.Compare(p.tagPairs[p.index].Name, p.mergeTags[p.mergeTagsIdx].Name) < 0 {
// the next provided tag is less
p.curr = p.tagPairs[p.index]
p.index++
} else {
// the next merge tag is less
p.curr = p.mergeTags[p.mergeTagsIdx]
p.mergeTagsIdx++
}
p.currIdx++
return true
}

func (p *rollupIDProvider) CurrentIndex() int {
if p.index >= 0 {
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this was actually broken before, so it's intentionally not backwards compatible. it would produce 0, 0, 1, 2, ...

the addition to the test proves this.

return p.index
}
return 0
return p.currIdx
}

func (p *rollupIDProvider) Current() ident.Tag {
idx := p.index
if idx == p.nameTagIndex {
p.tagValueID.Reset(p.newName)
return ident.Tag{
Name: p.nameTag,
Value: p.tagValueID,
}
}
if idx == p.rollupTagIndex {
return rollupTag
}

if p.index > p.nameTagIndex {
// Effective index is subtracted by 1
idx--
}
if p.index > p.rollupTagIndex {
// Effective index is subtracted by 1
idx--
}

p.tagNameID.Reset(p.tagPairs[idx].Name)
p.tagValueID.Reset(p.tagPairs[idx].Value)
p.tagNameID.Reset(p.curr.Name)
p.tagValueID.Reset(p.curr.Value)
return ident.Tag{
Name: p.tagNameID,
Value: p.tagValueID,
Expand All @@ -235,7 +217,7 @@ func (p *rollupIDProvider) Err() error {
func (p *rollupIDProvider) Close() {}

func (p *rollupIDProvider) Len() int {
return len(p.tagPairs) + 2
return p.len
}

func (p *rollupIDProvider) Remaining() int {
Expand Down
40 changes: 40 additions & 0 deletions src/cmd/services/m3coordinator/downsample/id_pool_types_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -267,6 +267,39 @@ func TestRollupIdProvider(t *testing.T) {
},
},
},
{
name: "rollup and name already exists",
nameTag: "__name__",
metricName: "http_requests",
tags: []id.TagPair{
{
Name: []byte("__name__"),
Value: []byte("http_requests"),
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

should we have a different value here to demonstrate which value is chosen?

},
{
Name: []byte("__rollup__"),
Value: []byte("true"),
},
{
Name: []byte("foo"),
Value: []byte("fooValue"),
},
},
expectedTags: []id.TagPair{
{
Name: []byte("__name__"),
Value: []byte("http_requests"),
},
{
Name: []byte("__rollup__"),
Value: []byte("true"),
},
{
Name: []byte("foo"),
Value: []byte("fooValue"),
},
},
},
}

for _, tc := range cases {
Expand All @@ -277,6 +310,13 @@ func TestRollupIdProvider(t *testing.T) {
}
encoder := &serialize.FakeTagEncoder{}
p := newRollupIDProvider(encoder, nil, ident.BytesID(tc.nameTag))
p.reset([]byte(tc.metricName), tc.tags)
require.Equal(t, len(tc.expectedTags), p.Len())
curIdx := 0
for p.Next() {
require.Equal(t, curIdx, p.CurrentIndex())
curIdx++
}
rollupID, err := p.provide([]byte(tc.metricName), tc.tags)
require.NoError(t, err)
encoded, _ := encoder.Data()
Expand Down