2013-02-07 10:49:04 +00:00
|
|
|
// Copyright 2013 Prometheus Team
|
|
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
|
|
// you may not use this file except in compliance with the License.
|
|
|
|
// You may obtain a copy of the License at
|
|
|
|
//
|
|
|
|
// http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
//
|
|
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
// See the License for the specific language governing permissions and
|
|
|
|
// limitations under the License.
|
|
|
|
|
2013-01-07 22:24:26 +00:00
|
|
|
package ast
|
|
|
|
|
|
|
|
import (
|
|
|
|
"errors"
|
2013-01-17 23:07:00 +00:00
|
|
|
"fmt"
|
2013-06-25 12:02:27 +00:00
|
|
|
"hash/fnv"
|
2013-01-07 22:24:26 +00:00
|
|
|
"math"
|
2013-01-17 23:07:00 +00:00
|
|
|
"sort"
|
2013-01-07 22:24:26 +00:00
|
|
|
"time"
|
2013-06-25 12:02:27 +00:00
|
|
|
|
2013-08-12 15:18:02 +00:00
|
|
|
"github.com/golang/glog"
|
|
|
|
|
2013-06-25 12:02:27 +00:00
|
|
|
clientmodel "github.com/prometheus/client_golang/model"
|
|
|
|
|
|
|
|
"github.com/prometheus/prometheus/stats"
|
|
|
|
"github.com/prometheus/prometheus/storage/metric"
|
2013-01-07 22:24:26 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------
|
|
|
|
// Raw data value types.
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Vector is basically only an alias for clientmodel.Samples, but the
|
|
|
|
// contract is that in a Vector, all Samples have the same timestamp.
|
2013-06-25 12:02:27 +00:00
|
|
|
type Vector clientmodel.Samples
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Matrix is a slice of SampleSets that implements sort.Interface and
|
|
|
|
// has a String method.
|
2013-06-25 12:02:27 +00:00
|
|
|
// BUG(julius): Pointerize this.
|
|
|
|
type Matrix []metric.SampleSet
|
2013-01-07 22:24:26 +00:00
|
|
|
|
|
|
|
type groupedAggregation struct {
|
2013-06-25 12:02:27 +00:00
|
|
|
labels clientmodel.Metric
|
|
|
|
value clientmodel.SampleValue
|
2013-01-07 22:24:26 +00:00
|
|
|
groupCount int
|
|
|
|
}
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------
|
|
|
|
// Enums.
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// ExprType is an enum for the rule language expression types.
|
2013-01-07 22:24:26 +00:00
|
|
|
type ExprType int
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Possible language expression types.
|
2013-01-07 22:24:26 +00:00
|
|
|
const (
|
|
|
|
SCALAR ExprType = iota
|
|
|
|
VECTOR
|
|
|
|
MATRIX
|
|
|
|
STRING
|
|
|
|
)
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// BinOpType is an enum for binary operator types.
|
2013-01-07 22:24:26 +00:00
|
|
|
type BinOpType int
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Possible binary operator types.
|
2013-01-07 22:24:26 +00:00
|
|
|
const (
|
|
|
|
ADD BinOpType = iota
|
|
|
|
SUB
|
|
|
|
MUL
|
|
|
|
DIV
|
|
|
|
MOD
|
|
|
|
NE
|
|
|
|
EQ
|
|
|
|
GT
|
|
|
|
LT
|
|
|
|
GE
|
|
|
|
LE
|
|
|
|
AND
|
|
|
|
OR
|
|
|
|
)
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// AggrType is an enum for aggregation types.
|
2013-01-07 22:24:26 +00:00
|
|
|
type AggrType int
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Possible aggregation types.
|
2013-01-07 22:24:26 +00:00
|
|
|
const (
|
|
|
|
SUM AggrType = iota
|
|
|
|
AVG
|
|
|
|
MIN
|
|
|
|
MAX
|
2013-05-08 14:35:16 +00:00
|
|
|
COUNT
|
2013-01-07 22:24:26 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------
|
|
|
|
// Interfaces.
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Nodes is a slice of any mix of node types as all node types
|
|
|
|
// implement the Node interface.
|
2013-06-06 13:12:37 +00:00
|
|
|
type Nodes []Node
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Node is the top-level interface for any kind of nodes. Each node
|
|
|
|
// type implements one of the ...Node interfaces, each of which embeds
|
|
|
|
// this Node interface.
|
2013-01-07 22:24:26 +00:00
|
|
|
type Node interface {
|
|
|
|
Type() ExprType
|
2013-06-06 13:12:37 +00:00
|
|
|
Children() Nodes
|
2013-01-07 22:24:26 +00:00
|
|
|
NodeTreeToDotGraph() string
|
2013-06-06 13:12:37 +00:00
|
|
|
String() string
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// ScalarNode is a Node for scalar values.
|
2013-01-07 22:24:26 +00:00
|
|
|
type ScalarNode interface {
|
|
|
|
Node
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval evaluates and returns the value of the scalar represented by this node.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
Eval(timestamp clientmodel.Timestamp, view *viewAdapter) clientmodel.SampleValue
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// VectorNode is a Node for vector values.
|
2013-01-07 22:24:26 +00:00
|
|
|
type VectorNode interface {
|
|
|
|
Node
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval evaluates the node recursively and returns the result
|
|
|
|
// as a Vector (i.e. a slice of Samples all at the given
|
|
|
|
// Timestamp).
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
Eval(timestamp clientmodel.Timestamp, view *viewAdapter) Vector
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// MatrixNode is a Node for matrix values.
|
2013-01-07 22:24:26 +00:00
|
|
|
type MatrixNode interface {
|
|
|
|
Node
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval evaluates the node recursively and returns the result as a Matrix.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
Eval(timestamp clientmodel.Timestamp, view *viewAdapter) Matrix
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval evaluates the node recursively and returns the result
|
|
|
|
// as a Matrix that only contains the boundary values.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
EvalBoundaries(timestamp clientmodel.Timestamp, view *viewAdapter) Matrix
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// StringNode is a Node for string values.
|
2013-01-07 22:24:26 +00:00
|
|
|
type StringNode interface {
|
|
|
|
Node
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval evaluates and returns the value of the string
|
|
|
|
// represented by this node.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
Eval(timestamp clientmodel.Timestamp, view *viewAdapter) string
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------
|
|
|
|
// ScalarNode types.
|
|
|
|
|
|
|
|
type (
|
2014-02-22 21:29:32 +00:00
|
|
|
// ScalarLiteral represents a numeric selector.
|
2013-01-07 22:24:26 +00:00
|
|
|
ScalarLiteral struct {
|
2013-06-25 12:02:27 +00:00
|
|
|
value clientmodel.SampleValue
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// ScalarFunctionCall represents a function with a numeric
|
|
|
|
// return type.
|
2013-01-07 22:24:26 +00:00
|
|
|
ScalarFunctionCall struct {
|
|
|
|
function *Function
|
2013-06-06 13:12:37 +00:00
|
|
|
args Nodes
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// ScalarArithExpr represents an arithmetic expression of
|
|
|
|
// numeric type.
|
2013-01-07 22:24:26 +00:00
|
|
|
ScalarArithExpr struct {
|
|
|
|
opType BinOpType
|
|
|
|
lhs ScalarNode
|
|
|
|
rhs ScalarNode
|
|
|
|
}
|
|
|
|
)
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------
|
|
|
|
// VectorNode types.
|
|
|
|
|
|
|
|
type (
|
2014-02-22 21:29:32 +00:00
|
|
|
// A VectorSelector represents a metric name plus labelset.
|
|
|
|
VectorSelector struct {
|
2013-06-25 12:02:27 +00:00
|
|
|
labels clientmodel.LabelSet
|
2013-03-27 13:06:30 +00:00
|
|
|
// Fingerprints are populated from labels at query analysis time.
|
2013-06-25 12:02:27 +00:00
|
|
|
fingerprints clientmodel.Fingerprints
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// VectorFunctionCall represents a function with vector return
|
|
|
|
// type.
|
2013-01-07 22:24:26 +00:00
|
|
|
VectorFunctionCall struct {
|
|
|
|
function *Function
|
2013-06-06 13:12:37 +00:00
|
|
|
args Nodes
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// A VectorAggregation with vector return type.
|
2013-01-07 22:24:26 +00:00
|
|
|
VectorAggregation struct {
|
2013-12-13 18:20:42 +00:00
|
|
|
aggrType AggrType
|
|
|
|
groupBy clientmodel.LabelNames
|
|
|
|
keepExtraLabels bool
|
|
|
|
vector VectorNode
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// VectorArithExpr represents an arithmetic expression of
|
|
|
|
// vector type.
|
2013-01-07 22:24:26 +00:00
|
|
|
VectorArithExpr struct {
|
|
|
|
opType BinOpType
|
|
|
|
lhs VectorNode
|
|
|
|
rhs Node
|
|
|
|
}
|
|
|
|
)
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------
|
|
|
|
// MatrixNode types.
|
|
|
|
|
|
|
|
type (
|
2014-02-22 21:29:32 +00:00
|
|
|
// A MatrixSelector represents a metric name plus labelset and
|
2014-02-13 17:48:56 +00:00
|
|
|
// timerange.
|
2014-02-22 21:29:32 +00:00
|
|
|
MatrixSelector struct {
|
2013-06-25 12:02:27 +00:00
|
|
|
labels clientmodel.LabelSet
|
2014-02-13 17:48:56 +00:00
|
|
|
// Fingerprints are populated from labels at query
|
|
|
|
// analysis time.
|
2013-06-25 12:02:27 +00:00
|
|
|
fingerprints clientmodel.Fingerprints
|
2013-03-26 11:33:48 +00:00
|
|
|
interval time.Duration
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
)
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------
|
|
|
|
// StringNode types.
|
|
|
|
|
|
|
|
type (
|
2014-02-13 17:48:56 +00:00
|
|
|
// A StringLiteral is what you think it is.
|
2013-01-07 22:24:26 +00:00
|
|
|
StringLiteral struct {
|
|
|
|
str string
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// StringFunctionCall represents a function with string return
|
|
|
|
// type.
|
2013-01-07 22:24:26 +00:00
|
|
|
StringFunctionCall struct {
|
|
|
|
function *Function
|
2013-06-06 13:12:37 +00:00
|
|
|
args Nodes
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
)
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------
|
|
|
|
// Implementations.
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Type implements the Node interface.
|
|
|
|
func (node ScalarLiteral) Type() ExprType { return SCALAR }
|
|
|
|
|
|
|
|
// Type implements the Node interface.
|
2013-01-07 22:24:26 +00:00
|
|
|
func (node ScalarFunctionCall) Type() ExprType { return SCALAR }
|
2014-02-13 17:48:56 +00:00
|
|
|
|
|
|
|
// Type implements the Node interface.
|
|
|
|
func (node ScalarArithExpr) Type() ExprType { return SCALAR }
|
|
|
|
|
|
|
|
// Type implements the Node interface.
|
2014-02-22 21:29:32 +00:00
|
|
|
func (node VectorSelector) Type() ExprType { return VECTOR }
|
2014-02-13 17:48:56 +00:00
|
|
|
|
|
|
|
// Type implements the Node interface.
|
2013-01-07 22:24:26 +00:00
|
|
|
func (node VectorFunctionCall) Type() ExprType { return VECTOR }
|
2014-02-13 17:48:56 +00:00
|
|
|
|
|
|
|
// Type implements the Node interface.
|
|
|
|
func (node VectorAggregation) Type() ExprType { return VECTOR }
|
|
|
|
|
|
|
|
// Type implements the Node interface.
|
|
|
|
func (node VectorArithExpr) Type() ExprType { return VECTOR }
|
|
|
|
|
|
|
|
// Type implements the Node interface.
|
2014-02-22 21:29:32 +00:00
|
|
|
func (node MatrixSelector) Type() ExprType { return MATRIX }
|
2014-02-13 17:48:56 +00:00
|
|
|
|
|
|
|
// Type implements the Node interface.
|
|
|
|
func (node StringLiteral) Type() ExprType { return STRING }
|
|
|
|
|
|
|
|
// Type implements the Node interface.
|
2013-01-07 22:24:26 +00:00
|
|
|
func (node StringFunctionCall) Type() ExprType { return STRING }
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Children implements the Node interface and returns an empty slice.
|
|
|
|
func (node ScalarLiteral) Children() Nodes { return Nodes{} }
|
|
|
|
|
|
|
|
// Children implements the Node interface and returns the args of the
|
|
|
|
// function call.
|
2013-06-06 13:12:37 +00:00
|
|
|
func (node ScalarFunctionCall) Children() Nodes { return node.args }
|
2014-02-13 17:48:56 +00:00
|
|
|
|
|
|
|
// Children implements the Node interface and returns the LHS and the RHS
|
|
|
|
// of the expression.
|
|
|
|
func (node ScalarArithExpr) Children() Nodes { return Nodes{node.lhs, node.rhs} }
|
|
|
|
|
|
|
|
// Children implements the Node interface and returns an empty slice.
|
2014-02-22 21:29:32 +00:00
|
|
|
func (node VectorSelector) Children() Nodes { return Nodes{} }
|
2014-02-13 17:48:56 +00:00
|
|
|
|
|
|
|
// Children implements the Node interface and returns the args of the
|
|
|
|
// function call.
|
2013-06-06 13:12:37 +00:00
|
|
|
func (node VectorFunctionCall) Children() Nodes { return node.args }
|
2014-02-13 17:48:56 +00:00
|
|
|
|
|
|
|
// Children implements the Node interface and returns the vector to be
|
|
|
|
// aggregated.
|
|
|
|
func (node VectorAggregation) Children() Nodes { return Nodes{node.vector} }
|
|
|
|
|
|
|
|
// Children implements the Node interface and returns the LHS and the RHS
|
|
|
|
// of the expression.
|
|
|
|
func (node VectorArithExpr) Children() Nodes { return Nodes{node.lhs, node.rhs} }
|
|
|
|
|
|
|
|
// Children implements the Node interface and returns an empty slice.
|
2014-02-22 21:29:32 +00:00
|
|
|
func (node MatrixSelector) Children() Nodes { return Nodes{} }
|
2014-02-13 17:48:56 +00:00
|
|
|
|
|
|
|
// Children implements the Node interface and returns an empty slice.
|
|
|
|
func (node StringLiteral) Children() Nodes { return Nodes{} }
|
|
|
|
|
|
|
|
// Children implements the Node interface and returns the args of the
|
|
|
|
// function call.
|
2013-06-06 13:12:37 +00:00
|
|
|
func (node StringFunctionCall) Children() Nodes { return node.args }
|
2013-03-21 17:06:15 +00:00
|
|
|
|
2014-02-22 21:29:32 +00:00
|
|
|
// Eval implements the ScalarNode interface and returns the selector
|
2014-02-13 17:48:56 +00:00
|
|
|
// value.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func (node *ScalarLiteral) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) clientmodel.SampleValue {
|
2013-01-07 22:24:26 +00:00
|
|
|
return node.value
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval implements the ScalarNode interface and returns the result of
|
|
|
|
// the expression.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func (node *ScalarArithExpr) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) clientmodel.SampleValue {
|
2013-03-21 17:06:15 +00:00
|
|
|
lhs := node.lhs.Eval(timestamp, view)
|
|
|
|
rhs := node.rhs.Eval(timestamp, view)
|
2013-01-07 22:24:26 +00:00
|
|
|
return evalScalarBinop(node.opType, lhs, rhs)
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval implements the ScalarNode interface and returns the result of
|
|
|
|
// the function call.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func (node *ScalarFunctionCall) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) clientmodel.SampleValue {
|
2013-06-25 12:02:27 +00:00
|
|
|
return node.function.callFn(timestamp, view, node.args).(clientmodel.SampleValue)
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2013-06-25 12:02:27 +00:00
|
|
|
func (node *VectorAggregation) labelsToGroupingKey(labels clientmodel.Metric) uint64 {
|
|
|
|
summer := fnv.New64a()
|
|
|
|
for _, label := range node.groupBy {
|
|
|
|
fmt.Fprint(summer, labels[label])
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
2013-06-25 12:02:27 +00:00
|
|
|
|
|
|
|
return summer.Sum64()
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2013-06-25 12:02:27 +00:00
|
|
|
func labelsToKey(labels clientmodel.Metric) uint64 {
|
|
|
|
pairs := metric.LabelPairs{}
|
|
|
|
|
2013-01-17 23:07:00 +00:00
|
|
|
for label, value := range labels {
|
2013-06-25 12:02:27 +00:00
|
|
|
pairs = append(pairs, &metric.LabelPair{
|
|
|
|
Name: label,
|
|
|
|
Value: value,
|
|
|
|
})
|
2013-01-17 23:07:00 +00:00
|
|
|
}
|
2013-06-25 12:02:27 +00:00
|
|
|
|
|
|
|
sort.Sort(pairs)
|
|
|
|
|
|
|
|
summer := fnv.New64a()
|
|
|
|
|
|
|
|
for _, pair := range pairs {
|
|
|
|
fmt.Fprint(summer, pair.Name, pair.Value)
|
|
|
|
}
|
|
|
|
|
|
|
|
return summer.Sum64()
|
2013-01-15 10:30:55 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// EvalVectorInstant evaluates a VectorNode with an instant query.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func EvalVectorInstant(node VectorNode, timestamp clientmodel.Timestamp, storage *metric.TieredStorage, queryStats *stats.TimerGroup) (vector Vector, err error) {
|
2013-06-03 15:07:03 +00:00
|
|
|
viewAdapter, err := viewAdapterForInstantQuery(node, timestamp, storage, queryStats)
|
2013-03-21 17:06:15 +00:00
|
|
|
if err != nil {
|
|
|
|
return
|
|
|
|
}
|
2013-04-09 11:47:20 +00:00
|
|
|
vector = node.Eval(timestamp, viewAdapter)
|
|
|
|
return
|
2013-03-21 17:06:15 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// EvalVectorRange evaluates a VectorNode with a range query.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func EvalVectorRange(node VectorNode, start clientmodel.Timestamp, end clientmodel.Timestamp, interval time.Duration, storage *metric.TieredStorage, queryStats *stats.TimerGroup) (Matrix, error) {
|
2013-03-25 11:14:48 +00:00
|
|
|
// Explicitly initialize to an empty matrix since a nil Matrix encodes to
|
|
|
|
// null in JSON.
|
2013-05-16 05:38:31 +00:00
|
|
|
matrix := Matrix{}
|
2013-03-25 11:14:48 +00:00
|
|
|
|
2013-06-03 15:07:03 +00:00
|
|
|
viewTimer := queryStats.GetTimer(stats.TotalViewBuildingTime).Start()
|
|
|
|
viewAdapter, err := viewAdapterForRangeQuery(node, start, end, interval, storage, queryStats)
|
|
|
|
viewTimer.Stop()
|
2013-03-21 17:06:15 +00:00
|
|
|
if err != nil {
|
2013-05-16 05:38:31 +00:00
|
|
|
return nil, err
|
2013-03-21 17:06:15 +00:00
|
|
|
}
|
2013-06-03 15:07:03 +00:00
|
|
|
|
2013-01-17 23:07:00 +00:00
|
|
|
// TODO implement watchdog timer for long-running queries.
|
2013-06-03 15:07:03 +00:00
|
|
|
evalTimer := queryStats.GetTimer(stats.InnerEvalTime).Start()
|
2013-06-25 12:02:27 +00:00
|
|
|
sampleSets := map[uint64]*metric.SampleSet{}
|
2013-03-21 17:06:15 +00:00
|
|
|
for t := start; t.Before(end); t = t.Add(interval) {
|
2013-03-28 16:05:06 +00:00
|
|
|
vector := node.Eval(t, viewAdapter)
|
2013-01-17 23:07:00 +00:00
|
|
|
for _, sample := range vector {
|
2013-06-25 12:02:27 +00:00
|
|
|
samplePair := &metric.SamplePair{
|
2013-01-17 23:07:00 +00:00
|
|
|
Value: sample.Value,
|
|
|
|
Timestamp: sample.Timestamp,
|
|
|
|
}
|
|
|
|
groupingKey := labelsToKey(sample.Metric)
|
|
|
|
if sampleSets[groupingKey] == nil {
|
2013-06-25 12:02:27 +00:00
|
|
|
sampleSets[groupingKey] = &metric.SampleSet{
|
2013-01-17 23:07:00 +00:00
|
|
|
Metric: sample.Metric,
|
2013-06-25 12:02:27 +00:00
|
|
|
Values: metric.Values{samplePair},
|
2013-01-17 23:07:00 +00:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
sampleSets[groupingKey].Values = append(sampleSets[groupingKey].Values, samplePair)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2013-06-03 15:07:03 +00:00
|
|
|
evalTimer.Stop()
|
2013-01-17 23:07:00 +00:00
|
|
|
|
2013-06-03 15:07:03 +00:00
|
|
|
appendTimer := queryStats.GetTimer(stats.ResultAppendTime).Start()
|
2013-01-17 23:07:00 +00:00
|
|
|
for _, sampleSet := range sampleSets {
|
2013-03-28 16:05:06 +00:00
|
|
|
matrix = append(matrix, *sampleSet)
|
2013-01-17 23:07:00 +00:00
|
|
|
}
|
2013-06-03 15:07:03 +00:00
|
|
|
appendTimer.Stop()
|
2013-05-16 05:38:31 +00:00
|
|
|
|
|
|
|
return matrix, nil
|
2013-01-15 10:30:55 +00:00
|
|
|
}
|
|
|
|
|
2013-06-25 12:02:27 +00:00
|
|
|
func labelIntersection(metric1, metric2 clientmodel.Metric) clientmodel.Metric {
|
|
|
|
intersection := clientmodel.Metric{}
|
2013-01-07 22:24:26 +00:00
|
|
|
for label, value := range metric1 {
|
|
|
|
if metric2[label] == value {
|
|
|
|
intersection[label] = value
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return intersection
|
|
|
|
}
|
|
|
|
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func (node *VectorAggregation) groupedAggregationsToVector(aggregations map[uint64]*groupedAggregation, timestamp clientmodel.Timestamp) Vector {
|
2013-01-07 22:24:26 +00:00
|
|
|
vector := Vector{}
|
|
|
|
for _, aggregation := range aggregations {
|
2013-05-08 14:35:16 +00:00
|
|
|
switch node.aggrType {
|
|
|
|
case AVG:
|
2013-06-25 12:02:27 +00:00
|
|
|
aggregation.value = aggregation.value / clientmodel.SampleValue(aggregation.groupCount)
|
2013-05-08 14:35:16 +00:00
|
|
|
case COUNT:
|
2013-06-25 12:02:27 +00:00
|
|
|
aggregation.value = clientmodel.SampleValue(aggregation.groupCount)
|
2013-05-08 14:35:16 +00:00
|
|
|
default:
|
|
|
|
// For other aggregations, we already have the right value.
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
2013-06-25 12:02:27 +00:00
|
|
|
sample := &clientmodel.Sample{
|
2013-01-07 22:24:26 +00:00
|
|
|
Metric: aggregation.labels,
|
|
|
|
Value: aggregation.value,
|
2013-03-28 16:05:06 +00:00
|
|
|
Timestamp: timestamp,
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
vector = append(vector, sample)
|
|
|
|
}
|
|
|
|
return vector
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval implements the VectorNode interface and returns the aggregated
|
|
|
|
// Vector.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func (node *VectorAggregation) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) Vector {
|
2013-03-21 17:06:15 +00:00
|
|
|
vector := node.vector.Eval(timestamp, view)
|
2013-06-25 12:02:27 +00:00
|
|
|
result := map[uint64]*groupedAggregation{}
|
2013-01-07 22:24:26 +00:00
|
|
|
for _, sample := range vector {
|
|
|
|
groupingKey := node.labelsToGroupingKey(sample.Metric)
|
|
|
|
if groupedResult, ok := result[groupingKey]; ok {
|
2013-12-13 18:20:42 +00:00
|
|
|
if node.keepExtraLabels {
|
|
|
|
groupedResult.labels = labelIntersection(groupedResult.labels, sample.Metric)
|
|
|
|
}
|
|
|
|
|
2013-01-07 22:24:26 +00:00
|
|
|
switch node.aggrType {
|
|
|
|
case SUM:
|
|
|
|
groupedResult.value += sample.Value
|
|
|
|
case AVG:
|
|
|
|
groupedResult.value += sample.Value
|
|
|
|
groupedResult.groupCount++
|
|
|
|
case MAX:
|
|
|
|
if groupedResult.value < sample.Value {
|
|
|
|
groupedResult.value = sample.Value
|
|
|
|
}
|
|
|
|
case MIN:
|
|
|
|
if groupedResult.value > sample.Value {
|
|
|
|
groupedResult.value = sample.Value
|
|
|
|
}
|
2013-05-08 14:35:16 +00:00
|
|
|
case COUNT:
|
|
|
|
groupedResult.groupCount++
|
|
|
|
default:
|
|
|
|
panic("Unknown aggregation type")
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
} else {
|
2013-12-13 18:20:42 +00:00
|
|
|
m := clientmodel.Metric{}
|
|
|
|
if node.keepExtraLabels {
|
|
|
|
m = sample.Metric
|
|
|
|
} else {
|
|
|
|
m[clientmodel.MetricNameLabel] = sample.Metric[clientmodel.MetricNameLabel]
|
|
|
|
for _, l := range node.groupBy {
|
2014-01-24 15:05:30 +00:00
|
|
|
if v, ok := sample.Metric[l]; ok {
|
|
|
|
m[l] = v
|
|
|
|
}
|
2013-12-13 18:20:42 +00:00
|
|
|
}
|
|
|
|
}
|
2013-01-07 22:24:26 +00:00
|
|
|
result[groupingKey] = &groupedAggregation{
|
2013-12-13 18:20:42 +00:00
|
|
|
labels: m,
|
2013-01-07 22:24:26 +00:00
|
|
|
value: sample.Value,
|
|
|
|
groupCount: 1,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2013-06-25 12:02:27 +00:00
|
|
|
|
2013-01-07 22:24:26 +00:00
|
|
|
return node.groupedAggregationsToVector(result, timestamp)
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval implements the VectorNode interface and returns the value of
|
2014-02-22 21:29:32 +00:00
|
|
|
// the selector.
|
|
|
|
func (node *VectorSelector) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) Vector {
|
2013-03-27 13:06:30 +00:00
|
|
|
values, err := view.GetValueAtTime(node.fingerprints, timestamp)
|
2013-01-07 22:24:26 +00:00
|
|
|
if err != nil {
|
2013-08-12 16:22:48 +00:00
|
|
|
glog.Error("Unable to get vector values: ", err)
|
2013-01-07 22:24:26 +00:00
|
|
|
return Vector{}
|
|
|
|
}
|
|
|
|
return values
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval implements the VectorNode interface and returns the result of
|
|
|
|
// the function call.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func (node *VectorFunctionCall) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) Vector {
|
2013-03-21 17:06:15 +00:00
|
|
|
return node.function.callFn(timestamp, view, node.args).(Vector)
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
func evalScalarBinop(opType BinOpType,
|
2013-06-25 12:02:27 +00:00
|
|
|
lhs clientmodel.SampleValue,
|
|
|
|
rhs clientmodel.SampleValue) clientmodel.SampleValue {
|
2013-01-07 22:24:26 +00:00
|
|
|
switch opType {
|
|
|
|
case ADD:
|
|
|
|
return lhs + rhs
|
|
|
|
case SUB:
|
|
|
|
return lhs - rhs
|
|
|
|
case MUL:
|
|
|
|
return lhs * rhs
|
|
|
|
case DIV:
|
|
|
|
if rhs != 0 {
|
|
|
|
return lhs / rhs
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return clientmodel.SampleValue(math.Inf(int(rhs)))
|
2013-01-07 22:24:26 +00:00
|
|
|
case MOD:
|
|
|
|
if rhs != 0 {
|
2013-06-25 12:02:27 +00:00
|
|
|
return clientmodel.SampleValue(int(lhs) % int(rhs))
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return clientmodel.SampleValue(math.Inf(int(rhs)))
|
2013-01-07 22:24:26 +00:00
|
|
|
case EQ:
|
|
|
|
if lhs == rhs {
|
|
|
|
return 1
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0
|
2013-01-07 22:24:26 +00:00
|
|
|
case NE:
|
|
|
|
if lhs != rhs {
|
|
|
|
return 1
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0
|
2013-01-07 22:24:26 +00:00
|
|
|
case GT:
|
|
|
|
if lhs > rhs {
|
|
|
|
return 1
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0
|
2013-01-07 22:24:26 +00:00
|
|
|
case LT:
|
|
|
|
if lhs < rhs {
|
|
|
|
return 1
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0
|
2013-01-07 22:24:26 +00:00
|
|
|
case GE:
|
|
|
|
if lhs >= rhs {
|
|
|
|
return 1
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0
|
2013-01-07 22:24:26 +00:00
|
|
|
case LE:
|
|
|
|
if lhs <= rhs {
|
|
|
|
return 1
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
panic("Not all enum values enumerated in switch")
|
|
|
|
}
|
|
|
|
|
|
|
|
func evalVectorBinop(opType BinOpType,
|
2013-06-25 12:02:27 +00:00
|
|
|
lhs clientmodel.SampleValue,
|
|
|
|
rhs clientmodel.SampleValue) (clientmodel.SampleValue, bool) {
|
2013-01-07 22:24:26 +00:00
|
|
|
switch opType {
|
|
|
|
case ADD:
|
|
|
|
return lhs + rhs, true
|
|
|
|
case SUB:
|
|
|
|
return lhs - rhs, true
|
|
|
|
case MUL:
|
|
|
|
return lhs * rhs, true
|
|
|
|
case DIV:
|
|
|
|
if rhs != 0 {
|
|
|
|
return lhs / rhs, true
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return clientmodel.SampleValue(math.Inf(int(rhs))), true
|
2013-01-07 22:24:26 +00:00
|
|
|
case MOD:
|
|
|
|
if rhs != 0 {
|
2013-06-25 12:02:27 +00:00
|
|
|
return clientmodel.SampleValue(int(lhs) % int(rhs)), true
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return clientmodel.SampleValue(math.Inf(int(rhs))), true
|
2013-01-07 22:24:26 +00:00
|
|
|
case EQ:
|
|
|
|
if lhs == rhs {
|
|
|
|
return lhs, true
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0, false
|
2013-01-07 22:24:26 +00:00
|
|
|
case NE:
|
|
|
|
if lhs != rhs {
|
|
|
|
return lhs, true
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0, false
|
2013-01-07 22:24:26 +00:00
|
|
|
case GT:
|
|
|
|
if lhs > rhs {
|
|
|
|
return lhs, true
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0, false
|
2013-01-07 22:24:26 +00:00
|
|
|
case LT:
|
|
|
|
if lhs < rhs {
|
|
|
|
return lhs, true
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0, false
|
2013-01-07 22:24:26 +00:00
|
|
|
case GE:
|
|
|
|
if lhs >= rhs {
|
|
|
|
return lhs, true
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0, false
|
2013-01-07 22:24:26 +00:00
|
|
|
case LE:
|
|
|
|
if lhs <= rhs {
|
|
|
|
return lhs, true
|
|
|
|
}
|
2014-02-13 17:48:56 +00:00
|
|
|
return 0, false
|
2013-01-07 22:24:26 +00:00
|
|
|
case AND:
|
|
|
|
return lhs, true
|
2013-01-17 22:45:18 +00:00
|
|
|
case OR:
|
|
|
|
return lhs, true // TODO: implement OR
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
panic("Not all enum values enumerated in switch")
|
|
|
|
}
|
|
|
|
|
2013-06-25 12:02:27 +00:00
|
|
|
func labelsEqual(labels1, labels2 clientmodel.Metric) bool {
|
2013-01-07 22:24:26 +00:00
|
|
|
if len(labels1) != len(labels2) {
|
|
|
|
return false
|
|
|
|
}
|
|
|
|
for label, value := range labels1 {
|
2013-06-25 12:02:27 +00:00
|
|
|
if labels2[label] != value && label != clientmodel.MetricNameLabel {
|
2013-01-07 22:24:26 +00:00
|
|
|
return false
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return true
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval implements the VectorNode interface and returns the result of
|
|
|
|
// the expression.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func (node *VectorArithExpr) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) Vector {
|
2013-03-21 17:06:15 +00:00
|
|
|
lhs := node.lhs.Eval(timestamp, view)
|
2013-01-07 22:24:26 +00:00
|
|
|
result := Vector{}
|
|
|
|
if node.rhs.Type() == SCALAR {
|
2013-03-21 17:06:15 +00:00
|
|
|
rhs := node.rhs.(ScalarNode).Eval(timestamp, view)
|
2013-01-07 22:24:26 +00:00
|
|
|
for _, lhsSample := range lhs {
|
|
|
|
value, keep := evalVectorBinop(node.opType, lhsSample.Value, rhs)
|
|
|
|
if keep {
|
|
|
|
lhsSample.Value = value
|
|
|
|
result = append(result, lhsSample)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return result
|
|
|
|
} else if node.rhs.Type() == VECTOR {
|
2013-03-21 17:06:15 +00:00
|
|
|
rhs := node.rhs.(VectorNode).Eval(timestamp, view)
|
2013-01-07 22:24:26 +00:00
|
|
|
for _, lhsSample := range lhs {
|
|
|
|
for _, rhsSample := range rhs {
|
|
|
|
if labelsEqual(lhsSample.Metric, rhsSample.Metric) {
|
|
|
|
value, keep := evalVectorBinop(node.opType, lhsSample.Value, rhsSample.Value)
|
|
|
|
if keep {
|
|
|
|
lhsSample.Value = value
|
|
|
|
result = append(result, lhsSample)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return result
|
|
|
|
}
|
|
|
|
panic("Invalid vector arithmetic expression operands")
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval implements the MatrixNode interface and returns the value of
|
2014-02-22 21:29:32 +00:00
|
|
|
// the selector.
|
|
|
|
func (node *MatrixSelector) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) Matrix {
|
2013-06-25 12:02:27 +00:00
|
|
|
interval := &metric.Interval{
|
2013-01-11 00:08:47 +00:00
|
|
|
OldestInclusive: timestamp.Add(-node.interval),
|
2013-03-28 16:05:06 +00:00
|
|
|
NewestInclusive: timestamp,
|
2013-01-11 00:08:47 +00:00
|
|
|
}
|
2013-03-27 13:06:30 +00:00
|
|
|
values, err := view.GetRangeValues(node.fingerprints, interval)
|
2013-01-07 22:24:26 +00:00
|
|
|
if err != nil {
|
2013-08-12 16:22:48 +00:00
|
|
|
glog.Error("Unable to get values for vector interval: ", err)
|
2013-01-07 22:24:26 +00:00
|
|
|
return Matrix{}
|
|
|
|
}
|
|
|
|
return values
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// EvalBoundaries implements the MatrixNode interface and returns the
|
2014-02-22 21:29:32 +00:00
|
|
|
// boundary values of the selector.
|
|
|
|
func (node *MatrixSelector) EvalBoundaries(timestamp clientmodel.Timestamp, view *viewAdapter) Matrix {
|
2013-06-25 12:02:27 +00:00
|
|
|
interval := &metric.Interval{
|
2013-01-07 22:24:26 +00:00
|
|
|
OldestInclusive: timestamp.Add(-node.interval),
|
2013-03-28 16:05:06 +00:00
|
|
|
NewestInclusive: timestamp,
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
2013-03-27 13:06:30 +00:00
|
|
|
values, err := view.GetBoundaryValues(node.fingerprints, interval)
|
2013-01-07 22:24:26 +00:00
|
|
|
if err != nil {
|
2013-08-12 16:22:48 +00:00
|
|
|
glog.Error("Unable to get boundary values for vector interval: ", err)
|
2013-01-07 22:24:26 +00:00
|
|
|
return Matrix{}
|
|
|
|
}
|
|
|
|
return values
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Len implements sort.Interface.
|
2013-01-15 10:30:55 +00:00
|
|
|
func (matrix Matrix) Len() int {
|
2013-01-17 23:07:00 +00:00
|
|
|
return len(matrix)
|
2013-01-15 10:30:55 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Less implements sort.Interface.
|
2013-01-15 10:30:55 +00:00
|
|
|
func (matrix Matrix) Less(i, j int) bool {
|
2013-06-25 12:02:27 +00:00
|
|
|
return matrix[i].Metric.String() < matrix[j].Metric.String()
|
2013-01-15 10:30:55 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Swap implements sort.Interface.
|
2013-01-15 10:30:55 +00:00
|
|
|
func (matrix Matrix) Swap(i, j int) {
|
2013-01-17 23:07:00 +00:00
|
|
|
matrix[i], matrix[j] = matrix[j], matrix[i]
|
2013-01-15 10:30:55 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval implements the StringNode interface and returns the value of
|
2014-02-22 21:29:32 +00:00
|
|
|
// the selector.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func (node *StringLiteral) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) string {
|
2013-01-07 22:24:26 +00:00
|
|
|
return node.str
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// Eval implements the StringNode interface and returns the result of
|
|
|
|
// the function call.
|
Use custom timestamp type for sample timestamps and related code.
So far we've been using Go's native time.Time for anything related to sample
timestamps. Since the range of time.Time is much bigger than what we need, this
has created two problems:
- there could be time.Time values which were out of the range/precision of the
time type that we persist to disk, therefore causing incorrectly ordered keys.
One bug caused by this was:
https://github.com/prometheus/prometheus/issues/367
It would be good to use a timestamp type that's more closely aligned with
what the underlying storage supports.
- sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit
Unix timestamp (possibly even a 32-bit one). Since we store samples in large
numbers, this seriously affects memory usage. Furthermore, copying/working
with the data will be faster if it's smaller.
*MEMORY USAGE RESULTS*
Initial memory usage comparisons for a running Prometheus with 1 timeseries and
100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my
tests, this advantage for some reason decreased a bit the more samples the
timeseries had (to 5-7% for millions of samples). This I can't fully explain,
but perhaps garbage collection issues were involved.
*WHEN TO USE THE NEW TIMESTAMP TYPE*
The new clientmodel.Timestamp type should be used whenever time
calculations are either directly or indirectly related to sample
timestamps.
For example:
- the timestamp of a sample itself
- all kinds of watermarks
- anything that may become or is compared to a sample timestamp (like the timestamp
passed into Target.Scrape()).
When to still use time.Time:
- for measuring durations/times not related to sample timestamps, like duration
telemetry exporting, timers that indicate how frequently to execute some
action, etc.
*NOTE ON OPERATOR OPTIMIZATION TESTS*
We don't use operator optimization code anymore, but it still lives in
the code as dead code. It still has tests, but I couldn't get all of them to
pass with the new timestamp format. I commented out the failing cases for now,
but we should probably remove the dead code soon. I just didn't want to do that
in the same change as this.
Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
|
|
|
func (node *StringFunctionCall) Eval(timestamp clientmodel.Timestamp, view *viewAdapter) string {
|
2013-03-21 17:06:15 +00:00
|
|
|
return node.function.callFn(timestamp, view, node.args).(string)
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------
|
|
|
|
// Constructors.
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// NewScalarLiteral returns a ScalarLiteral with the given value.
|
2013-06-25 12:02:27 +00:00
|
|
|
func NewScalarLiteral(value clientmodel.SampleValue) *ScalarLiteral {
|
2013-01-07 22:24:26 +00:00
|
|
|
return &ScalarLiteral{
|
|
|
|
value: value,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-02-22 21:29:32 +00:00
|
|
|
// NewVectorSelector returns a (not yet evaluated) VectorSelector with
|
2014-02-13 17:48:56 +00:00
|
|
|
// the given LabelSet.
|
2014-02-22 21:29:32 +00:00
|
|
|
func NewVectorSelector(labels clientmodel.LabelSet) *VectorSelector {
|
|
|
|
return &VectorSelector{
|
2013-01-07 22:24:26 +00:00
|
|
|
labels: labels,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// NewVectorAggregation returns a (not yet evaluated)
|
|
|
|
// VectorAggregation, aggregating the given VectorNode using the given
|
|
|
|
// AggrType, grouping by the given LabelNames.
|
2013-12-13 18:20:42 +00:00
|
|
|
func NewVectorAggregation(aggrType AggrType, vector VectorNode, groupBy clientmodel.LabelNames, keepExtraLabels bool) *VectorAggregation {
|
2013-01-07 22:24:26 +00:00
|
|
|
return &VectorAggregation{
|
2013-12-13 18:20:42 +00:00
|
|
|
aggrType: aggrType,
|
|
|
|
groupBy: groupBy,
|
|
|
|
keepExtraLabels: keepExtraLabels,
|
|
|
|
vector: vector,
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// NewFunctionCall returns a (not yet evaluated) function call node
|
|
|
|
// (of type ScalarFunctionCall, VectorFunctionCall, or
|
|
|
|
// StringFunctionCall).
|
2013-06-06 13:12:37 +00:00
|
|
|
func NewFunctionCall(function *Function, args Nodes) (Node, error) {
|
2013-01-07 22:24:26 +00:00
|
|
|
if err := function.CheckArgTypes(args); err != nil {
|
|
|
|
return nil, err
|
|
|
|
}
|
|
|
|
switch function.returnType {
|
|
|
|
case SCALAR:
|
|
|
|
return &ScalarFunctionCall{
|
|
|
|
function: function,
|
|
|
|
args: args,
|
|
|
|
}, nil
|
|
|
|
case VECTOR:
|
|
|
|
return &VectorFunctionCall{
|
|
|
|
function: function,
|
|
|
|
args: args,
|
|
|
|
}, nil
|
|
|
|
case STRING:
|
|
|
|
return &StringFunctionCall{
|
|
|
|
function: function,
|
|
|
|
args: args,
|
|
|
|
}, nil
|
|
|
|
}
|
|
|
|
panic("Function with invalid return type")
|
|
|
|
}
|
|
|
|
|
2013-06-06 13:12:37 +00:00
|
|
|
func nodesHaveTypes(nodes Nodes, exprTypes []ExprType) bool {
|
2013-01-07 22:24:26 +00:00
|
|
|
for _, node := range nodes {
|
2013-01-12 20:22:59 +00:00
|
|
|
correctType := false
|
2013-01-07 22:24:26 +00:00
|
|
|
for _, exprType := range exprTypes {
|
|
|
|
if node.Type() == exprType {
|
2013-01-11 00:09:31 +00:00
|
|
|
correctType = true
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
}
|
2013-01-12 20:22:59 +00:00
|
|
|
if !correctType {
|
|
|
|
return false
|
|
|
|
}
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
2013-01-11 00:09:31 +00:00
|
|
|
return true
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// NewArithExpr returns a (not yet evaluated) expression node (of type
|
|
|
|
// VectorArithExpr or ScalarArithExpr).
|
2013-01-07 22:24:26 +00:00
|
|
|
func NewArithExpr(opType BinOpType, lhs Node, rhs Node) (Node, error) {
|
2013-06-06 13:12:37 +00:00
|
|
|
if !nodesHaveTypes(Nodes{lhs, rhs}, []ExprType{SCALAR, VECTOR}) {
|
2014-02-13 17:48:56 +00:00
|
|
|
return nil, errors.New("binary operands must be of vector or scalar type")
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
if lhs.Type() == SCALAR && rhs.Type() == VECTOR {
|
2014-02-13 17:48:56 +00:00
|
|
|
return nil, errors.New("left side of vector binary operation must be of vector type")
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if opType == AND || opType == OR {
|
|
|
|
if lhs.Type() == SCALAR || rhs.Type() == SCALAR {
|
|
|
|
return nil, errors.New("AND and OR operators may only be used between vectors")
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if lhs.Type() == VECTOR || rhs.Type() == VECTOR {
|
|
|
|
return &VectorArithExpr{
|
|
|
|
opType: opType,
|
|
|
|
lhs: lhs.(VectorNode),
|
|
|
|
rhs: rhs,
|
|
|
|
}, nil
|
|
|
|
}
|
|
|
|
|
|
|
|
return &ScalarArithExpr{
|
|
|
|
opType: opType,
|
|
|
|
lhs: lhs.(ScalarNode),
|
|
|
|
rhs: rhs.(ScalarNode),
|
|
|
|
}, nil
|
|
|
|
}
|
|
|
|
|
2014-02-22 21:29:32 +00:00
|
|
|
// NewMatrixSelector returns a (not yet evaluated) MatrixSelector with
|
|
|
|
// the given VectorSelector and Duration.
|
|
|
|
func NewMatrixSelector(vector *VectorSelector, interval time.Duration) *MatrixSelector {
|
|
|
|
return &MatrixSelector{
|
2013-01-07 22:24:26 +00:00
|
|
|
labels: vector.labels,
|
|
|
|
interval: interval,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-02-13 17:48:56 +00:00
|
|
|
// NewStringLiteral returns a StringLiteral with the given string as
|
|
|
|
// value.
|
2013-01-07 22:24:26 +00:00
|
|
|
func NewStringLiteral(str string) *StringLiteral {
|
|
|
|
return &StringLiteral{
|
|
|
|
str: str,
|
|
|
|
}
|
|
|
|
}
|