logs-analyzer/signoz/pkg/query-service/app/logparsingpipeline/postablePipeline.go
2024-09-02 22:47:30 +03:00

239 lines
7.5 KiB
Go

package logparsingpipeline
import (
"errors"
"fmt"
"regexp"
"strings"
v3 "go.signoz.io/signoz/pkg/query-service/model/v3"
"go.signoz.io/signoz/pkg/query-service/queryBuilderToExpr"
"golang.org/x/exp/slices"
)
// PostablePipelines are a list of user defined pielines
type PostablePipelines struct {
Pipelines []PostablePipeline `json:"pipelines"`
}
// PostablePipeline captures user inputs in setting the pipeline
type PostablePipeline struct {
Id string `json:"id"`
OrderId int `json:"orderId"`
Name string `json:"name"`
Alias string `json:"alias"`
Description string `json:"description"`
Enabled bool `json:"enabled"`
Filter *v3.FilterSet `json:"filter"`
Config []PipelineOperator `json:"config"`
}
// IsValid checks if postable pipeline has all the required params
func (p *PostablePipeline) IsValid() error {
if p.OrderId == 0 {
return fmt.Errorf("orderId with value > 1 is required")
}
if p.Name == "" {
return fmt.Errorf("pipeline name is required")
}
if p.Alias == "" {
return fmt.Errorf("pipeline alias is required")
}
// check the filter
_, err := queryBuilderToExpr.Parse(p.Filter)
if err != nil {
return fmt.Errorf(fmt.Sprintf("filter for pipeline %v is not correct: %v", p.Name, err.Error()))
}
idUnique := map[string]struct{}{}
outputUnique := map[string]struct{}{}
l := len(p.Config)
for i, op := range p.Config {
if op.OrderId == 0 {
return fmt.Errorf("orderId with value > 1 is required in operator")
}
if op.ID == "" {
return fmt.Errorf("id of an operator cannot be empty")
}
if op.Type == "" {
return fmt.Errorf("type of an operator cannot be empty")
}
if i != (l-1) && op.Output == "" {
return fmt.Errorf(fmt.Sprintf("Output of operator %s cannot be nil", op.ID))
}
if i == (l-1) && op.Output != "" {
return fmt.Errorf(fmt.Sprintf("Output of operator %s should be empty", op.ID))
}
if _, ok := idUnique[op.ID]; ok {
return fmt.Errorf("duplicate id cannot be present")
}
if _, ok := outputUnique[op.Output]; ok {
return fmt.Errorf("duplicate output cannot be present")
}
if op.ID == op.Output {
return fmt.Errorf("id and output cannot be same")
}
err := isValidOperator(op)
if err != nil {
return err
}
idUnique[op.ID] = struct{}{}
outputUnique[op.Output] = struct{}{}
}
return nil
}
func isValidOperator(op PipelineOperator) error {
if op.ID == "" {
return errors.New("PipelineOperator.ID is required")
}
switch op.Type {
case "json_parser":
if op.ParseFrom == "" && op.ParseTo == "" {
return fmt.Errorf(fmt.Sprintf("parse from and parse to of %s json operator cannot be empty", op.ID))
}
case "grok_parser":
if op.Pattern == "" {
return fmt.Errorf(fmt.Sprintf("pattern of %s grok operator cannot be empty", op.ID))
}
case "regex_parser":
if op.Regex == "" {
return fmt.Errorf(fmt.Sprintf("regex of %s regex operator cannot be empty", op.ID))
}
r, err := regexp.Compile(op.Regex)
if err != nil {
return fmt.Errorf(fmt.Sprintf("error compiling regex expression of %s regex operator", op.ID))
}
namedCaptureGroups := 0
for _, groupName := range r.SubexpNames() {
if groupName != "" {
namedCaptureGroups++
}
}
if namedCaptureGroups == 0 {
return fmt.Errorf(fmt.Sprintf("no capture groups in regex expression of %s regex operator", op.ID))
}
case "copy":
if op.From == "" || op.To == "" {
return fmt.Errorf(fmt.Sprintf("from or to of %s copy operator cannot be empty", op.ID))
}
case "move":
if op.From == "" || op.To == "" {
return fmt.Errorf(fmt.Sprintf("from or to of %s move operator cannot be empty", op.ID))
}
case "add":
if op.Field == "" || op.Value == "" {
return fmt.Errorf(fmt.Sprintf("field or value of %s add operator cannot be empty", op.ID))
}
case "remove":
if op.Field == "" {
return fmt.Errorf(fmt.Sprintf("field of %s remove operator cannot be empty", op.ID))
}
case "trace_parser":
if op.TraceParser == nil {
return fmt.Errorf(fmt.Sprintf("field of %s remove operator cannot be empty", op.ID))
}
hasTraceIdParseFrom := (op.TraceParser.TraceId != nil && op.TraceParser.TraceId.ParseFrom != "")
hasSpanIdParseFrom := (op.TraceParser.SpanId != nil && op.TraceParser.SpanId.ParseFrom != "")
hasTraceFlagsParseFrom := (op.TraceParser.TraceFlags != nil && op.TraceParser.TraceFlags.ParseFrom != "")
if !(hasTraceIdParseFrom || hasSpanIdParseFrom || hasTraceFlagsParseFrom) {
return fmt.Errorf(fmt.Sprintf("one of trace_id, span_id, trace_flags of %s trace_parser operator must be present", op.ID))
}
if hasTraceIdParseFrom && !isValidOtelValue(op.TraceParser.TraceId.ParseFrom) {
return fmt.Errorf("trace id can't be parsed from %s", op.TraceParser.TraceId.ParseFrom)
}
if hasSpanIdParseFrom && !isValidOtelValue(op.TraceParser.SpanId.ParseFrom) {
return fmt.Errorf("span id can't be parsed from %s", op.TraceParser.SpanId.ParseFrom)
}
if hasTraceFlagsParseFrom && !isValidOtelValue(op.TraceParser.TraceFlags.ParseFrom) {
return fmt.Errorf("trace flags can't be parsed from %s", op.TraceParser.TraceFlags.ParseFrom)
}
case "retain":
if len(op.Fields) == 0 {
return fmt.Errorf(fmt.Sprintf("fields of %s retain operator cannot be empty", op.ID))
}
case "time_parser":
if op.ParseFrom == "" {
return fmt.Errorf("parse from of time parsing processor %s cannot be empty", op.ID)
}
if op.LayoutType != "epoch" && op.LayoutType != "strptime" {
// TODO(Raj): Maybe add support for gotime format
return fmt.Errorf(
"invalid format type '%s' of time parsing processor %s", op.LayoutType, op.ID,
)
}
if op.Layout == "" {
return fmt.Errorf(fmt.Sprintf("format can not be empty for time parsing processor %s", op.ID))
}
validEpochLayouts := []string{"s", "ms", "us", "ns", "s.ms", "s.us", "s.ns"}
if op.LayoutType == "epoch" && !slices.Contains(validEpochLayouts, op.Layout) {
return fmt.Errorf(
"invalid epoch format '%s' of time parsing processor %s", op.LayoutType, op.ID,
)
}
// TODO(Raj): Add validation for strptime layouts via
// collector simulator maybe.
if op.LayoutType == "strptime" {
_, err := RegexForStrptimeLayout(op.Layout)
if err != nil {
return fmt.Errorf(
"invalid strptime format '%s' of time parsing processor %s: %w", op.LayoutType, op.ID, err,
)
}
}
case "severity_parser":
if op.ParseFrom == "" {
return fmt.Errorf("parse from of severity parsing processor %s cannot be empty", op.ID)
}
validMappingLevels := []string{"trace", "debug", "info", "warn", "error", "fatal"}
for k := range op.SeverityMapping {
if !slices.Contains(validMappingLevels, strings.ToLower(k)) {
return fmt.Errorf("%s is not a valid severity in processor %s", k, op.ID)
}
}
default:
return fmt.Errorf(fmt.Sprintf("operator type %s not supported for %s, use one of (grok_parser, regex_parser, copy, move, add, remove, trace_parser, retain)", op.Type, op.ID))
}
if !isValidOtelValue(op.ParseFrom) ||
!isValidOtelValue(op.ParseTo) ||
!isValidOtelValue(op.From) ||
!isValidOtelValue(op.To) ||
!isValidOtelValue(op.Field) {
valueErrStr := "value should have prefix of body, attributes, resource"
return fmt.Errorf(fmt.Sprintf("%s for operator Id %s", valueErrStr, op.ID))
}
return nil
}
func isValidOtelValue(val string) bool {
if val == "" {
return true
}
if !strings.HasPrefix(val, "body") &&
!strings.HasPrefix(val, "attributes") &&
!strings.HasPrefix(val, "resource") {
return false
}
return true
}