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

lightning: optimize region split check logic #30428

Merged
merged 5 commits into from
Dec 20, 2021
Merged

lightning: optimize region split check logic #30428

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
040d68a
optimize region split check log and lower the split keys threshold
glorv Dec 6, 2021
c1d257c
adjust default region-max-keys
glorv Dec 6, 2021
c22f395
Merge branch 'master' into opt-region-split
glorv Dec 6, 2021
3d30fff
Merge branch 'master' into opt-region-split
ti-chi-bot Dec 20, 2021
d7e67bb
Merge branch 'master' into opt-region-split
ti-chi-bot Dec 20, 2021
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
10 changes: 8 additions & 2 deletions br/pkg/lightning/backend/local/local.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,7 +82,8 @@ const (
gRPCBackOffMaxDelay = 10 * time.Minute

// See: https://github.com/tikv/tikv/blob/e030a0aae9622f3774df89c62f21b2171a72a69e/etc/config-template.toml#L360
regionMaxKeyCount = 1_440_000
// lower the max-key-count to avoid tikv trigger region auto split
regionMaxKeyCount = 1_280_000
defaultRegionSplitSize = 96 * units.MiB

propRangeIndex = "tikv.range_index"
Expand Down Expand Up @@ -782,7 +783,12 @@ func (local *local) WriteToTiKV(
size := int64(0)
totalCount := int64(0)
firstLoop := true
regionMaxSize := regionSplitSize * 4 / 3
// if region-split-size <= 96MiB, we bump the threshold a bit to avoid too many retry split
// because the range-properties is not 100% accurate
Copy link
Contributor

Choose a reason for hiding this comment

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

Can we desrease defaultPropSizeIndexDistance to get more accurate results? Since it doesn't cost too much memory.

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 have tested smaller value 1MiB/10000, but there is no much difference. Maybe we can do deeper investigation to know the root reason.😅

regionMaxSize := regionSplitSize
if regionSplitSize <= defaultRegionSplitSize {
regionMaxSize = regionSplitSize * 4 / 3
}

for iter.First(); iter.Valid(); iter.Next() {
size += int64(len(iter.Key()) + len(iter.Value()))
Expand Down
4 changes: 2 additions & 2 deletions br/pkg/lightning/restore/table_restore.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -998,8 +998,8 @@ func estimateCompactionThreshold(cp *checkpoints.TableCheckpoint, factor int64)
threshold := totalRawFileSize / 512
threshold = utils.NextPowerOfTwo(threshold)
if threshold < compactionLowerThreshold {
// disable compaction if threshold is smaller than lower bound
threshold = 0
// too may small SST files will cause inaccuracy of region range estimation,
threshold = compactionLowerThreshold
} else if threshold > compactionUpperThreshold {
threshold = compactionUpperThreshold
}
Expand Down