Fix panic: runtime error: slice bounds out of range

[jwilder]

This fixes a regression in 1.4 introduced in ca40c1a where a delete with an invalid time range could end up either causing a panic or duplicate data in the blocks. I discovered this trying to reproduce #9116, which I have not been able to reproduce yet. This fix may fix that as well. This does not occur in 1.3.

For both of these, star the server with INFLUXDB_DATA_CACHE_SNAPSHOT_WRITE_COLD_DURATION=1s influxd

To reproduce the panic:

A delete where the min and max times are backwards causes a panic in Exclude.

#!/bin/bash

influx -execute "drop database test; create database test"

curl -v "http://localhost:8086/write?db=test" --data-binary 'cpu value=1 0
cpu value=1 1
cpu value=1 2
cpu value=1 3
cpu value=1 4
cpu value=1 5
cpu value=1 6
cpu value=1 7
cpu value=1 8
cpu value=1 9
cpu value=1 10
cpu value=1 11
mem value=1 1
zzz value=1 2
aaa value=1 3'

echo "Waiting 2 seconds for snapshot"
sleep 2

echo "All data is there"
influx -database test -execute "select * from /.*/"

echo "Delete cpu series by exact times"
influx -database test -execute "delete from cpu where time >= 10 and time <= 1;"


echo "Waiting for compaction"
sleep 2
echo "All series lexicographically after aaa are gone"

influx -database test -execute "select * from /.*/"

The server crashes with:

panic: runtime error: slice bounds out of range

goroutine 4248 [running]:
github.com/influxdata/influxdb/tsdb/engine/tsm1.FloatValues.Exclude(0xc42031e000, 0xc, 0xc, 0xa, 0x1, 0xc42031e000, 0xc, 0xc)
	/Users/jason/go/src/github.com/influxdata/influxdb/tsdb/engine/tsm1/encoding.gen.go:317 +0x1c8
github.com/influxdata/influxdb/tsdb/engine/tsm1.(*tsmKeyIterator).combineFloat(0xc4204d31e0, 0xc4202ba101, 0x8, 0xc4207457d0, 0x10452d4)
	/Users/jason/go/src/github.com/influxdata/influxdb/tsdb/engine/tsm1/compact.gen.go:87 +0x430
github.com/influxdata/influxdb/tsdb/engine/tsm1.(*tsmKeyIterator).mergeFloat(0xc4204d31e0)
	/Users/jason/go/src/github.com/influxdata/influxdb/tsdb/engine/tsm1/compact.gen.go:31 +0x14e

To reproduce the duplicate data:

#!/bin/bash

influx -execute "drop database test; create database test"

curl -v "http://localhost:8086/write?db=test" --data-binary 'cpu value=1 0
cpu value=1 1
cpu value=1 2
cpu value=1 3
cpu value=1 4
cpu value=1 5
cpu value=1 6
cpu value=1 7
cpu value=1 8
cpu value=1 9
cpu value=1 10
cpu value=1 11
mem value=1 1
zzz value=1 2
aaa value=1 3'

echo "Waiting 2 seconds for snapshot"
sleep 2

echo "All data is there"
influx -database test -execute "select * from /.*/"

echo "Delete cpu series by exact times"
influx -database test -execute "delete from cpu where time >= 0 and time <= 0; delete from cpu where time >= 2 and time <= 4; delete from cpu where time >= 6 and time <= 8; delete from cpu where time >= 10 and time <= 1;"

echo "Waiting for compaction"
sleep 2
echo "All series lexicographically after aaa are gone"

influx -database test -execute "select * from /.*/"

This causes the CPU series to have duplicate data after a compaction:

name: cpu
time value
---- -----
1    1
5    1
9    1
5    1
9    1
10   1
11   1

Required for all non-trivial PRs

• Rebased/mergable
• Tests pass
• CHANGELOG.md updated
• Sign CLA (if not already signed)

[e-dard]

LGTM 👍

[stuartcarnie]

LGTM!