Drop all disruption/alert matcher fallbacks except previous release

These turned out to be failing 100% of the time and doing all kinds of
things we wouldn't expect. If we don't have the data, we won't
run the test.

Author: dgoodwin

pkg/monitortestlibrary/allowedalerts/matches_test.go

@@ -106,23 +106,6 @@ func TestGetClosestP99Value(t *testing.T) {
 			},
 			expectedDuration: mustDuration("7.9s"),
 		},
-		{
-			name: "choose different arch",
-			key: historicaldata.AlertDataKey{
-				AlertName:      "etcdGRPCRequestsSlow",
-				AlertNamespace: "",
-				AlertLevel:     "warning",
-				JobType: platformidentification.JobType{
-					Release:      "4.12",
-					FromRelease:  "4.12",
-					Platform:     "aws",
-					Architecture: "not-real",
-					Network:      "sdn",
-					Topology:     "ha",
-				},
-			},
-			expectedDuration: mustDuration("120.458s"),
-		},
 		{
 			name: "missing",
 			key: historicaldata.AlertDataKey{

pkg/monitortestlibrary/allowedbackenddisruption/matches_test.go

@@ -152,53 +152,6 @@ func TestGetClosestP95Value(t *testing.T) {
 			},
 			expectedDuration: mustDuration("2.987s"),
 		},
-		{
-			name: "fuzzy match single ovn fallback to sdn",
-			args: args{
-				backendName: "kube-api-new-connections",
-				jobType: platformidentification.JobType{
-					Release:      "4.12",
-					FromRelease:  "4.12",
-					Platform:     "aws",
-					Architecture: "amd64",
-					Network:      "ovn", // we only defined sdn above, should fall back to it's value
-					Topology:     "single",
-				},
-			},
-			expectedDuration: mustDuration("120.458s"),
-		},
-		{
-			name: "fuzzy match single ovn fallback to sdn previous",
-			args: args{
-				// switching to openshift-api backend, defined above we only have from 4.11->4.11 and sdn
-				backendName: "openshift-api-new-connections",
-				jobType: platformidentification.JobType{
-					Release:      "4.12",
-					FromRelease:  "4.12",
-					Platform:     "aws",
-					Network:      "ovn",
-					Architecture: "amd64",
-					Topology:     "single",
-				},
-			},
-			expectedDuration: mustDuration("70.381s"),
-		},
-		{
-			name: "fuzzy match single ovn fallback to sdn previous micro",
-			args: args{
-				// For oauth backend ovn single don't have 4.12 minor, but we have 4.11->4.12 sdn above:
-				backendName: "oauth-api-new-connections",
-				jobType: platformidentification.JobType{
-					Release:      "4.12",
-					FromRelease:  "4.11",
-					Platform:     "aws",
-					Network:      "ovn",
-					Architecture: "amd64",
-					Topology:     "single",
-				},
-			},
-			expectedDuration: mustDuration("35.917s"),
-		},
 		{
 			name: "no exact or fuzzy match",
 			args: args{

pkg/monitortestlibrary/historicaldata/next_best_guess.go

@@ -10,95 +10,18 @@ import (
 )
 
 // nextBestGuessers is the order in which to attempt to lookup other alternative matches that are close to this job type.
-// TODO building a cross multiply would likely be beneficial
 var nextBestGuessers = []NextBestKey{
-	MicroReleaseUpgrade,
-	//	MinorReleaseUpgrade,
+	// The only guesser we try not is falling back to previous release. Otherwise if we don't have enough data, we don't
+	// run the test. This was implemented after finding that we fail every attempt at a fallback.
+	// Continuing with previous release helps us in the transition between major releases, so we kept this fallback.
 	PreviousReleaseUpgrade,
-
-	combine(PreviousReleaseUpgrade, MicroReleaseUpgrade),
-	//	combine(PreviousReleaseUpgrade, MinorReleaseUpgrade),
-
-	OnArchitecture("amd64"),
-	OnArchitecture("ppc64le"),
-	OnArchitecture("s390x"),
-	OnArchitecture("arm64"),
-
-	combine(OnArchitecture("amd64"), MicroReleaseUpgrade),
-	combine(OnArchitecture("ppc64le"), MicroReleaseUpgrade),
-	combine(OnArchitecture("s390x"), MicroReleaseUpgrade),
-	combine(OnArchitecture("arm64"), MicroReleaseUpgrade),
-
-	//	combine(OnArchitecture("amd64"), MinorReleaseUpgrade),
-	//	combine(OnArchitecture("ppc64le"), MinorReleaseUpgrade),
-	//	combine(OnArchitecture("s390x"), MinorReleaseUpgrade),
-	//	combine(OnArchitecture("arm64"), MinorReleaseUpgrade),
-
-	combine(OnArchitecture("amd64"), PreviousReleaseUpgrade),
-	combine(OnArchitecture("ppc64le"), PreviousReleaseUpgrade),
-	combine(OnArchitecture("s390x"), PreviousReleaseUpgrade),
-	combine(OnArchitecture("arm64"), PreviousReleaseUpgrade),
-
-	combine(OnArchitecture("amd64"), PreviousReleaseUpgrade, MicroReleaseUpgrade),
-	combine(OnArchitecture("ppc64le"), PreviousReleaseUpgrade, MicroReleaseUpgrade),
-	combine(OnArchitecture("s390x"), PreviousReleaseUpgrade, MicroReleaseUpgrade),
-	combine(OnArchitecture("arm64"), PreviousReleaseUpgrade, MicroReleaseUpgrade),
-
-	//	combine(OnArchitecture("amd64"), PreviousReleaseUpgrade, MinorReleaseUpgrade),
-	//	combine(OnArchitecture("ppc64le"), PreviousReleaseUpgrade, MinorReleaseUpgrade),
-	//	combine(OnArchitecture("s390x"), PreviousReleaseUpgrade, MinorReleaseUpgrade),
-	//	combine(OnArchitecture("arm64"), PreviousReleaseUpgrade, MinorReleaseUpgrade),
-
-	combine(ForTopology("single"), OnSDN),
-	combine(ForTopology("single"), OnSDN, PreviousReleaseUpgrade),
-	combine(ForTopology("single"), OnSDN, PreviousReleaseUpgrade, MicroReleaseUpgrade),
 }
 
 // NextBestKey returns the next best key in the query_results.json generated from BigQuery and a bool indicating whether this guesser has an opinion.
 // If the bool is false, the key should not be used.
 // Returning true doesn't mean the key exists, it just means that the key is worth trying.
 type NextBestKey func(in platformidentification.JobType) (platformidentification.JobType, bool)
 
-// MinorReleaseUpgrade if we don't have data for the current fromRelease and it's a micro upgrade, perhaps we have data
-// for a minor upgrade.  A 4.11 to 4.11 upgrade will attempt a 4.10 to 4.11 upgrade.
-func MinorReleaseUpgrade(in platformidentification.JobType) (platformidentification.JobType, bool) {
-	if len(in.FromRelease) == 0 {
-		return platformidentification.JobType{}, false
-	}
-
-	fromReleaseMinor := getMinor(in.FromRelease)
-	toReleaseMajor := getMajor(in.Release)
-	toReleaseMinor := getMinor(in.Release)
-	// if we're already a minor upgrade, this doesn't apply
-	if fromReleaseMinor == (toReleaseMinor - 1) {
-		return platformidentification.JobType{}, false
-	}
-
-	ret := platformidentification.CloneJobType(in)
-	ret.FromRelease = fmt.Sprintf("%d.%d", toReleaseMajor, toReleaseMinor-1)
-	return ret, true
-}
-
-// MicroReleaseUpgrade if we don't have data for the current fromRelease and it's a minor upgrade, perhaps we have data
-// for a micro upgrade.  A 4.10 to 4.11 upgrade will attempt a 4.11 to 4.11 upgrade.
-func MicroReleaseUpgrade(in platformidentification.JobType) (platformidentification.JobType, bool) {
-	if len(in.FromRelease) == 0 {
-		return platformidentification.JobType{}, false
-	}
-
-	fromReleaseMinor := getMinor(in.FromRelease)
-	toReleaseMajor := getMajor(in.Release)
-	toReleaseMinor := getMinor(in.Release)
-	// if we're already a micro upgrade, this doesn't apply
-	if fromReleaseMinor == toReleaseMinor {
-		return platformidentification.JobType{}, false
-	}
-
-	ret := platformidentification.CloneJobType(in)
-	ret.FromRelease = fmt.Sprintf("%d.%d", toReleaseMajor, toReleaseMinor)
-	return ret, true
-}
-
 // PreviousReleaseUpgrade if we don't have data for the current toRelease, perhaps we have data for the congruent test
 // on the prior release.   A 4.11 to 4.11 upgrade will attempt a 4.10 to 4.10 upgrade.  A 4.11 no upgrade, will attempt a 4.10 no upgrade.
 func PreviousReleaseUpgrade(in platformidentification.JobType) (platformidentification.JobType, bool) {
@@ -130,68 +53,6 @@ func getMinor(in string) int {
 	return int(minor)
 }
 
-// OnOVN maybe we have data on OVN
-func OnOVN(in platformidentification.JobType) (platformidentification.JobType, bool) {
-	if in.Network == "ovn" {
-		return platformidentification.JobType{}, false
-	}
-
-	ret := platformidentification.CloneJobType(in)
-	ret.Network = "ovn"
-	return ret, true
-}
-
-// OnSDN maybe we have data on SDN
-func OnSDN(in platformidentification.JobType) (platformidentification.JobType, bool) {
-	if in.Network == "sdn" {
-		return platformidentification.JobType{}, false
-	}
-
-	ret := platformidentification.CloneJobType(in)
-	ret.Network = "sdn"
-	return ret, true
-}
-
-// OnArchitecture maybe we match a different architecture
-func OnArchitecture(architecture string) func(in platformidentification.JobType) (platformidentification.JobType, bool) {
-	return func(in platformidentification.JobType) (platformidentification.JobType, bool) {
-		if in.Architecture == architecture {
-			return platformidentification.JobType{}, false
-		}
-
-		ret := platformidentification.CloneJobType(in)
-		ret.Architecture = architecture
-		return ret, true
-	}
-}
-
-// ForTopology we match on exact topology
-func ForTopology(topology string) func(in platformidentification.JobType) (platformidentification.JobType, bool) {
-	return func(in platformidentification.JobType) (platformidentification.JobType, bool) {
-		if in.Topology != topology {
-			return platformidentification.JobType{}, false
-		}
-		return in, true
-	}
-}
-
-// combine will start with the input and call each guess in order.  It uses the output of the previous NextBestKeyFn
-// as the input to the next.  This allows combinations like "previous release upgrade micro" without writing custom
-// functions for each.
-func combine(nextBestKeys ...NextBestKey) NextBestKey {
-	return func(in platformidentification.JobType) (platformidentification.JobType, bool) {
-		curr := in
-		for _, nextBestKey := range nextBestKeys {
-			var ok bool
-			curr, ok = nextBestKey(curr)
-			if !ok {
-				return curr, false
-			}
-		}
-		return curr, true
-	}
-}
-
 func CurrentReleaseFromMap(releasesInQueryResults map[string]bool) string {
 	var releaseSlice []struct {
 		Key   string