roachprod: implement GetVMSpecs for IBM

Implementation for previously unimplemented interface method `GetVMSpecs()` for
IBM. This will allow for cluster information to be fetched in roachtest via
`FetchVMSpecs`.

Informs #146202

Epic: None
Release note: None

Author: williamchoe3

pkg/roachprod/vm/ibm/BUILD.bazel

@@ -18,6 +18,7 @@ go_library(
         "//pkg/roachprod/logger",
         "//pkg/roachprod/vm",
         "//pkg/roachprod/vm/flagstub",
+        "//pkg/util/ctxgroup",
         "//pkg/util/randutil",
         "//pkg/util/retry",
         "//pkg/util/syncutil",

pkg/roachprod/vm/ibm/account_config.go

@@ -121,7 +121,7 @@ func (p *Provider) configureCloudAccountFull(cc *ibmCloudConfig) error {
 	// Create a transit gateway for the account in the first region.
 	tgID, err := p.getOrCreateTransitGateway(supportedRegions[0])
 	if err != nil {
-		return errors.Wrap(err, "failed to create transit gateway")
+		return errors.Wrap(err, "failed to get or create transit gateway")
 	}
 
 	for region := range p.regions {

pkg/roachprod/vm/ibm/helpers.go

@@ -1259,16 +1259,23 @@ type parsedCRN struct {
 	id     string
 }
 
+// parseCRN takes an IBM Cloud Resource Name string and parses it into
+// its components for easier use.
+// We don't store the resource ID in vm.VM, we need to parse the CRN if
+// we need the resource ID which is at index 9 after splitting.
+// CRN format:
+//
+//	crn:version:cname:ctype:service-name:location:scope:service-instance:resource-type:resource
+//
+// CRN e.g.
+//
+//	crn:v1:bluemix:public:is:ca-tor-1:a/ba0325a6257b4dabb3a7aa149a6578ae::instance:02q7_1a2940f7-b058-4742-8ef1-b1ee08273cf0
 func (p *Provider) parseCRN(crn string) (*parsedCRN, error) {
 
 	if crn == "" {
 		return nil, fmt.Errorf("empty CRN")
 	}
 
-	// CRN have the following format:
-	// crn:version:cname:ctype:service-name:location:scope:service-instance:resource-type:resource
-	// e.g.: crn:v1:bluemix:public:is:ca-tor-1:a/ba0325a6257b4dabb3a7aa149a6578ae::instance:02q7_1a2940f7-b058-4742-8ef1-b1ee08273cf0
-	// we're interested in the "resource" part, which is the 10th part
 	parts := strings.Split(crn, ":")
 
 	if len(parts) < 10 {
@@ -1279,6 +1286,9 @@ func (p *Provider) parseCRN(crn string) (*parsedCRN, error) {
 		id: parts[9],
 	}
 
+	// a CRN's location field can be a region or a zone, handle both cases
+	// region e.g. br-sao
+	// zone e.g. br-sao-1
 	splitsRegion := strings.Split(parts[5], "-")
 	if len(splitsRegion) < 3 {
 		c.region = parts[5]

pkg/roachprod/vm/ibm/provider.go

@@ -6,6 +6,8 @@
 package ibm
 
 import (
+	"context"
+	"encoding/json"
 	"fmt"
 	"os"
 	"os/exec"
@@ -23,6 +25,7 @@ import (
 	"github.com/cockroachdb/cockroach/pkg/roachprod/logger"
 	"github.com/cockroachdb/cockroach/pkg/roachprod/vm"
 	"github.com/cockroachdb/cockroach/pkg/roachprod/vm/flagstub"
+	"github.com/cockroachdb/cockroach/pkg/util/ctxgroup"
 	"github.com/cockroachdb/cockroach/pkg/util/randutil"
 	"github.com/cockroachdb/cockroach/pkg/util/syncutil"
 	"github.com/cockroachdb/cockroach/pkg/util/timeutil"
@@ -216,13 +219,29 @@ func NewProvider(options ...Option) (p *Provider, err error) {
 
 	// If an authenticator is not provided, create one from the environment.
 	if p.authenticator == nil {
+		// GetAuthenticatorFromEnvironment returns nil if IBM_APIKEY isn't set
+		// 1) First tries to find a credential file
+		//		a) using IBM_CREDENTIALS_FILE which is not set
+		//		b) <user-home-dir>/ibm-credentials.env which is not currently used
+		//		c) <current-working-directory>/ibm-credentials.env which is not currently used
+		// 2) Then uses the passed in string argument as a prefix and matches to
+		//	  any env var with a matching prefix. If IBM_APIKEY isn't set then
+		//	  this function continues.
+		// 3) Finally, will try to load VCAP_SERVICES which is not currently used
+		//
+		// If all these fail to match, GetAuthenticatorFromEnvironment returns nil
+		// without an error. We need to fail in that case, otherwise we will
+		// continue on with a nil p.authenticator which will cause failures later
 		p.authenticator, err = core.GetAuthenticatorFromEnvironment(defaultAPIKeyEnvVarPrefix)
 		if err != nil {
 			return nil, errors.Wrapf(
 				err,
 				"failed to create default IBM authenticator from environment variables",
 			)
 		}
+		if p.authenticator == nil {
+			return nil, errors.Newf("failed to create IBM authenticator. Is %s set?", expectedEnvVarIBMAPIKey)
+		}
 	}
 
 	// Build the list of supported regions if none are provided.
@@ -864,6 +883,70 @@ func (p *Provider) FindActiveAccount(l *logger.Logger) (string, error) {
 	return username, nil
 }
 
+// GetVMSpecs implements the vm.GetVMSpecs interface method which returns a
+// map from VM.Name to a map of VM attributes
+func (p *Provider) GetVMSpecs(
+	l *logger.Logger, vms vm.List,
+) (map[string]map[string]interface{}, error) {
+	// TODO replace background context with passed in context after *WithContext
+	// interface methods are implemented for vm.Provider
+	return p.GetVMSpecsWithContext(context.Background(), l, vms)
+}
+
+func (p *Provider) GetVMSpecsWithContext(
+	ctx context.Context, l *logger.Logger, vms vm.List,
+) (map[string]map[string]interface{}, error) {
+
+	// Extract the spec of all VMs and create a map from VM name to spec.
+	vmSpecs := make(map[string]map[string]interface{})
+	var mu syncutil.Mutex
+	g := ctxgroup.WithContext(ctx)
+	g.SetLimit(5)
+
+	for _, vmInstance := range vms {
+		g.GoCtx(func(ctx context.Context) error {
+			perCtx, cancel := context.WithTimeout(ctx, time.Minute)
+			defer cancel()
+
+			region, err := p.zoneToRegion(vmInstance.Zone)
+			if err != nil {
+				return err
+			}
+			vpcService, err := p.getVpcService(region)
+			if err != nil {
+				return err
+			}
+			parsedCrn, err := p.parseCRN(vmInstance.ProviderID)
+			if err != nil {
+				return err
+			}
+			l.Printf("Fetching specs for instance %s", vmInstance.Name)
+			result, _, err := vpcService.GetInstanceWithContext(perCtx, vpcService.NewGetInstanceOptions(parsedCrn.id))
+			if err != nil {
+				return err
+			}
+			// Marshaling & unmarshalling struct to match interface method return type
+			jsonBytes, err := json.Marshal(result)
+			if err != nil {
+				return err
+			}
+			var vmSpec map[string]interface{}
+			err = json.Unmarshal(jsonBytes, &vmSpec)
+			if err != nil {
+				return err
+			}
+			mu.Lock()
+			vmSpecs[vmInstance.Name] = vmSpec
+			mu.Unlock()
+			return nil
+		})
+	}
+	if err := g.Wait(); err != nil {
+		return nil, err
+	}
+	return vmSpecs, nil
+}
+
 //
 // Unimplemented methods, no plans to implement as of now.
 //
@@ -885,13 +968,6 @@ func (p *Provider) GetLiveMigrationVMs(
 	return nil, nil
 }
 
-// GetVMSpecs is part of the vm.Provider interface.
-func (p *Provider) GetVMSpecs(
-	_ *logger.Logger, _ vm.List,
-) (map[string]map[string]interface{}, error) {
-	return nil, nil
-}
-
 // Grow is part of the vm.Provider interface.
 func (p *Provider) Grow(_ *logger.Logger, _ vm.List, _ string, _ []string) (vm.List, error) {
 	return nil, vm.UnimplementedError