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.

Sign up for GitHub

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

Fix rewrite chunk reuse reference #3710

Closed
wants to merge 1 commit into from
Closed

Fix rewrite chunk reuse reference #3710

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
7 changes: 4 additions & 3 deletions pkg/compactv2/chunk_series_set.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,16 +47,17 @@ func (s *lazyPopulateChunkSeriesSet) Next() bool {
continue
}

chks := make([]chunks.Meta, len(s.bufChks))
Copy link
Member

Choose a reason for hiding this comment

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

Nice, so this is allowing us to not care about locking, but instead allocate more mem, right? 🤗

How benchmarks are looking with this?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Sorry I didn't get the point. How is it related to locking? Did you mean adding a lock to protect s.bufChks?

I haven't benchmarked this. For a one-off job, memory seems not the highest priority. Maybe we can call GC manually somewhere if the mem usage is too high?

I am not sure if we can fix this without allocating additional memory. What do you think?

Copy link
Member

Choose a reason for hiding this comment

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

I mean, no lock is needed because we copy the slice?

Copy link
Member

Choose a reason for hiding this comment

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

let's talk offline (:

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Sorry but I think I missed something. Is s.bufChks accessed by multiple goroutines? I thought it is accessed sequentially.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Not yet. 🤣

Copy link
Member

Choose a reason for hiding this comment

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

;p

Copy link
Member

Choose a reason for hiding this comment

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

Let's restart this. Can you explain what problem are you solving? you mentioned in the description that you are fixing some problem, which one? (:

Just creating slices everywhere is not the best idea if we care about efficiency here, we never seen any error around this place on production too.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I cannot tell tbh. But with this change the relabel pr works. I guess it is the problem of that pr because the rewrite delete works perfect.
But cannot figure out why.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I cannot tell tbh. But with this change the relabel pr works. I guess it is the problem of that pr because the rewrite delete works perfect.
But cannot figure out why.

copy(chks, s.bufChks)
for i := range s.bufChks {
s.bufChks[i].Chunk = &lazyPopulatableChunk{cr: s.sReader.cr, m: &s.bufChks[i]}
chks[i].Chunk = &lazyPopulatableChunk{cr: s.sReader.cr, m: &chks[i]}
}
s.curr = &storage.ChunkSeriesEntry{
Lset: make(labels.Labels, len(s.bufLbls)),
ChunkIteratorFn: func() chunks.Iterator {
return storage.NewListChunkSeriesIterator(s.bufChks...)
return storage.NewListChunkSeriesIterator(chks...)
},
}
// TODO: Do we need to copy this?
copy(s.curr.Lset, s.bufLbls)
return true
}
Expand Down
18 changes: 14 additions & 4 deletions pkg/compactv2/modifiers.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,8 @@
package compactv2

import (
"math"

"github.com/pkg/errors"
"github.com/prometheus/prometheus/storage"
"github.com/prometheus/prometheus/tsdb"
Expand Down Expand Up @@ -77,17 +79,25 @@ SeriesLoop:

// Special case: Delete whole series.
chksIter := s.Iterator()
var chks []chunks.Meta
var minT int64 = math.MaxInt64
var maxT int64 = math.MinInt64
for chksIter.Next() {
chks = append(chks, chksIter.At())
c := chksIter.At()
if c.MinTime < minT {
minT = c.MinTime
}
if c.MaxTime > maxT {
maxT = c.MaxTime
}
}
if d.err = chksIter.Err(); d.err != nil {
return false
}

var deleted tombstones.Intervals
if len(chks) > 0 {
deleted = deleted.Add(tombstones.Interval{Mint: chks[0].MinTime, Maxt: chks[len(chks)-1].MaxTime})
// If minTime is set then there is at least one chunk.
if minT != math.MaxInt64 {
deleted = deleted.Add(tombstones.Interval{Mint: minT, Maxt: maxT})
}
d.log.DeleteSeries(lbls, deleted)
d.p.SeriesProcessed()
Expand Down