20755: Allow cloud drivers to register their own metrics.

Arvados-DCO-1.1-Signed-off-by: Tom Clegg <tom@curii.com>

diff --git a/lib/cloud/azure/azure.go b/lib/cloud/azure/azure.go
index 494db854ed3c857fbf47f714f46b5c605120e30e..71f2a23dc963baa747f173d446135175754ffeec 100644 (file)
--- a/lib/cloud/azure/azure.go
+++ b/lib/cloud/azure/azure.go
@@ -28,6 +28,7 @@ import (
        "github.com/Azure/go-autorest/autorest/azure/auth"
        "github.com/Azure/go-autorest/autorest/to"
        "github.com/jmcvetta/randutil"
+       "github.com/prometheus/client_golang/prometheus"
        "github.com/sirupsen/logrus"
        "golang.org/x/crypto/ssh"
 )
@@ -238,7 +239,7 @@ type azureInstanceSet struct {
        logger             logrus.FieldLogger
 }
 
-func newAzureInstanceSet(config json.RawMessage, dispatcherID cloud.InstanceSetID, _ cloud.SharedResourceTags, logger logrus.FieldLogger) (prv cloud.InstanceSet, err error) {
+func newAzureInstanceSet(config json.RawMessage, dispatcherID cloud.InstanceSetID, _ cloud.SharedResourceTags, logger logrus.FieldLogger, reg *prometheus.Registry) (prv cloud.InstanceSet, err error) {
        azcfg := azureInstanceSetConfig{}
        err = json.Unmarshal(config, &azcfg)
        if err != nil {

diff --git a/lib/cloud/azure/azure_test.go b/lib/cloud/azure/azure_test.go
index 2f88f7344044e6e31d80e9ef00469771a0d0a116..de8d655b197a685a6b9a6124a63652fa1bb6a527 100644 (file)
--- a/lib/cloud/azure/azure_test.go
+++ b/lib/cloud/azure/azure_test.go
@@ -156,7 +156,7 @@ func GetInstanceSet() (*azureInstanceSet, cloud.ImageID, arvados.Cluster, error)
                        return nil, cloud.ImageID(""), cluster, err
                }
 
-               ap, err := newAzureInstanceSet(exampleCfg.DriverParameters, "test123", nil, logrus.StandardLogger())
+               ap, err := newAzureInstanceSet(exampleCfg.DriverParameters, "test123", nil, logrus.StandardLogger(), nil)
                return ap.(*azureInstanceSet), cloud.ImageID(exampleCfg.ImageIDForTestSuite), cluster, err
        }
        ap := azureInstanceSet{

diff --git a/lib/cloud/cloudtest/tester.go b/lib/cloud/cloudtest/tester.go
index 41e8f658ac20f606b1cc89ea2426bb474813e13b..d53cb017edfd910171b2f6f75de0f5feaedf5381 100644 (file)
--- a/lib/cloud/cloudtest/tester.go
+++ b/lib/cloud/cloudtest/tester.go
@@ -64,7 +64,7 @@ func (t *tester) Run() bool {
        deferredError := false
 
        var err error
-       t.is, err = t.Driver.InstanceSet(t.DriverParameters, t.SetID, t.Tags, t.Logger)
+       t.is, err = t.Driver.InstanceSet(t.DriverParameters, t.SetID, t.Tags, t.Logger, nil)
        if err != nil {
                t.Logger.WithError(err).Info("error initializing driver")
                return false

diff --git a/lib/cloud/ec2/ec2.go b/lib/cloud/ec2/ec2.go
index 526fc1307dc29e8d77d588cf6dae4c9489c0464d..b98e519d0a4a7852bf64db398ad984ec91373580 100644 (file)
--- a/lib/cloud/ec2/ec2.go
+++ b/lib/cloud/ec2/ec2.go
@@ -29,6 +29,7 @@ import (
        "github.com/aws/aws-sdk-go/aws/request"
        "github.com/aws/aws-sdk-go/aws/session"
        "github.com/aws/aws-sdk-go/service/ec2"
+       "github.com/prometheus/client_golang/prometheus"
        "github.com/sirupsen/logrus"
        "golang.org/x/crypto/ssh"
 )
@@ -114,7 +115,7 @@ type ec2InstanceSet struct {
        pricesUpdated map[priceKey]time.Time
 }
 
-func newEC2InstanceSet(config json.RawMessage, instanceSetID cloud.InstanceSetID, _ cloud.SharedResourceTags, logger logrus.FieldLogger) (prv cloud.InstanceSet, err error) {
+func newEC2InstanceSet(config json.RawMessage, instanceSetID cloud.InstanceSetID, _ cloud.SharedResourceTags, logger logrus.FieldLogger, reg *prometheus.Registry) (prv cloud.InstanceSet, err error) {
        instanceSet := &ec2InstanceSet{
                instanceSetID: instanceSetID,
                logger:        logger,

diff --git a/lib/cloud/ec2/ec2_test.go b/lib/cloud/ec2/ec2_test.go
index 2f3d319e0be3b706e4fbdaef0000bdeba8972fc4..d8322e7af38ab145e9a23b4e01554dae44fc6760 100644 (file)
--- a/lib/cloud/ec2/ec2_test.go
+++ b/lib/cloud/ec2/ec2_test.go
@@ -246,7 +246,7 @@ func GetInstanceSet(c *check.C) (*ec2InstanceSet, cloud.ImageID, arvados.Cluster
                err := config.LoadFile(&exampleCfg, *live)
                c.Assert(err, check.IsNil)
 
-               ap, err := newEC2InstanceSet(exampleCfg.DriverParameters, "test123", nil, logrus.StandardLogger())
+               ap, err := newEC2InstanceSet(exampleCfg.DriverParameters, "test123", nil, logrus.StandardLogger(), nil)
                c.Assert(err, check.IsNil)
                return ap.(*ec2InstanceSet), cloud.ImageID(exampleCfg.ImageIDForTestSuite), cluster
        }

diff --git a/lib/cloud/interfaces.go b/lib/cloud/interfaces.go
index 27cf26152c2962fa1b8d9afeba34f4fad57e2922..7c532fda4a9e6beaa1811ae3205dc592f186c315 100644 (file)
--- a/lib/cloud/interfaces.go
+++ b/lib/cloud/interfaces.go
@@ -11,6 +11,7 @@ import (
        "time"
 
        "git.arvados.org/arvados.git/sdk/go/arvados"
+       "github.com/prometheus/client_golang/prometheus"
        "github.com/sirupsen/logrus"
        "golang.org/x/crypto/ssh"
 )
@@ -191,7 +192,7 @@ type InitCommand string
 //
 //     type exampleDriver struct {}
 //
-//     func (*exampleDriver) InstanceSet(config json.RawMessage, id cloud.InstanceSetID, tags cloud.SharedResourceTags, logger logrus.FieldLogger) (cloud.InstanceSet, error) {
+//     func (*exampleDriver) InstanceSet(config json.RawMessage, id cloud.InstanceSetID, tags cloud.SharedResourceTags, logger logrus.FieldLogger, reg *prometheus.Registry) (cloud.InstanceSet, error) {
 //             var is exampleInstanceSet
 //             if err := json.Unmarshal(config, &is); err != nil {
 //                     return nil, err
@@ -199,20 +200,18 @@ type InitCommand string
 //             is.ownID = id
 //             return &is, nil
 //     }
-//
-//     var _ = registerCloudDriver("example", &exampleDriver{})
 type Driver interface {
-       InstanceSet(config json.RawMessage, id InstanceSetID, tags SharedResourceTags, logger logrus.FieldLogger) (InstanceSet, error)
+       InstanceSet(config json.RawMessage, id InstanceSetID, tags SharedResourceTags, logger logrus.FieldLogger, reg *prometheus.Registry) (InstanceSet, error)
 }
 
 // DriverFunc makes a Driver using the provided function as its
 // InstanceSet method. This is similar to http.HandlerFunc.
-func DriverFunc(fn func(config json.RawMessage, id InstanceSetID, tags SharedResourceTags, logger logrus.FieldLogger) (InstanceSet, error)) Driver {
+func DriverFunc(fn func(config json.RawMessage, id InstanceSetID, tags SharedResourceTags, logger logrus.FieldLogger, reg *prometheus.Registry) (InstanceSet, error)) Driver {
        return driverFunc(fn)
 }
 
-type driverFunc func(config json.RawMessage, id InstanceSetID, tags SharedResourceTags, logger logrus.FieldLogger) (InstanceSet, error)
+type driverFunc func(config json.RawMessage, id InstanceSetID, tags SharedResourceTags, logger logrus.FieldLogger, reg *prometheus.Registry) (InstanceSet, error)
 
-func (df driverFunc) InstanceSet(config json.RawMessage, id InstanceSetID, tags SharedResourceTags, logger logrus.FieldLogger) (InstanceSet, error) {
-       return df(config, id, tags, logger)
+func (df driverFunc) InstanceSet(config json.RawMessage, id InstanceSetID, tags SharedResourceTags, logger logrus.FieldLogger, reg *prometheus.Registry) (InstanceSet, error) {
+       return df(config, id, tags, logger, reg)
 }

diff --git a/lib/cloud/loopback/loopback.go b/lib/cloud/loopback/loopback.go
index 8afaa452570c1b79c90989edd8247408f5eb2b54..41878acd2285f56946dd94b978371bc24f7380a7 100644 (file)
--- a/lib/cloud/loopback/loopback.go
+++ b/lib/cloud/loopback/loopback.go
@@ -21,6 +21,7 @@ import (
        "git.arvados.org/arvados.git/lib/cloud"
        "git.arvados.org/arvados.git/lib/dispatchcloud/test"
        "git.arvados.org/arvados.git/sdk/go/arvados"
+       "github.com/prometheus/client_golang/prometheus"
        "github.com/sirupsen/logrus"
        "golang.org/x/crypto/ssh"
 )
@@ -45,7 +46,7 @@ type instanceSet struct {
        mtx           sync.Mutex
 }
 
-func newInstanceSet(config json.RawMessage, instanceSetID cloud.InstanceSetID, _ cloud.SharedResourceTags, logger logrus.FieldLogger) (cloud.InstanceSet, error) {
+func newInstanceSet(config json.RawMessage, instanceSetID cloud.InstanceSetID, _ cloud.SharedResourceTags, logger logrus.FieldLogger, reg *prometheus.Registry) (cloud.InstanceSet, error) {
        is := &instanceSet{
                instanceSetID: instanceSetID,
                logger:        logger,

diff --git a/lib/cloud/loopback/loopback_test.go b/lib/cloud/loopback/loopback_test.go
index 5c30f5f0e1b24e9de092316a318eff08b67c2800..0716179cb7c335b600a024f4627a108a7d8eec15 100644 (file)
--- a/lib/cloud/loopback/loopback_test.go
+++ b/lib/cloud/loopback/loopback_test.go
@@ -29,7 +29,7 @@ var _ = check.Suite(&suite{})
 
 func (*suite) TestCreateListExecDestroy(c *check.C) {
        logger := ctxlog.TestLogger(c)
-       is, err := Driver.InstanceSet(json.RawMessage("{}"), "testInstanceSetID", cloud.SharedResourceTags{"sharedTag": "sharedTagValue"}, logger)
+       is, err := Driver.InstanceSet(json.RawMessage("{}"), "testInstanceSetID", cloud.SharedResourceTags{"sharedTag": "sharedTagValue"}, logger, nil)
        c.Assert(err, check.IsNil)
 
        clientRSAKey, err := rsa.GenerateKey(rand.Reader, 1024)

diff --git a/lib/dispatchcloud/driver.go b/lib/dispatchcloud/driver.go
index 93515defb7d8ebb68c1a5800770c203d52aeaa38..44adc23fd37c28b660755ab81426a03fd0716c08 100644 (file)
--- a/lib/dispatchcloud/driver.go
+++ b/lib/dispatchcloud/driver.go
@@ -33,7 +33,7 @@ func newInstanceSet(cluster *arvados.Cluster, setID cloud.InstanceSetID, logger
                return nil, fmt.Errorf("unsupported cloud driver %q", cluster.Containers.CloudVMs.Driver)
        }
        sharedResourceTags := cloud.SharedResourceTags(cluster.Containers.CloudVMs.ResourceTags)
-       is, err := driver.InstanceSet(cluster.Containers.CloudVMs.DriverParameters, setID, sharedResourceTags, logger)
+       is, err := driver.InstanceSet(cluster.Containers.CloudVMs.DriverParameters, setID, sharedResourceTags, logger, reg)
        is = newInstrumentedInstanceSet(is, reg)
        if maxops := cluster.Containers.CloudVMs.MaxCloudOpsPerSecond; maxops > 0 {
                is = rateLimitedInstanceSet{

diff --git a/lib/dispatchcloud/test/stub_driver.go b/lib/dispatchcloud/test/stub_driver.go
index e91878527c45f82f934b8968e1f208f308088d33..037580f8d924429794fd968e2590a104035fc7bf 100644 (file)
--- a/lib/dispatchcloud/test/stub_driver.go
+++ b/lib/dispatchcloud/test/stub_driver.go
@@ -20,6 +20,7 @@ import (
        "git.arvados.org/arvados.git/lib/cloud"
        "git.arvados.org/arvados.git/lib/crunchrun"
        "git.arvados.org/arvados.git/sdk/go/arvados"
+       "github.com/prometheus/client_golang/prometheus"
        "github.com/sirupsen/logrus"
        "golang.org/x/crypto/ssh"
 )
@@ -62,7 +63,7 @@ type StubDriver struct {
 }
 
 // InstanceSet returns a new *StubInstanceSet.
-func (sd *StubDriver) InstanceSet(params json.RawMessage, id cloud.InstanceSetID, _ cloud.SharedResourceTags, logger logrus.FieldLogger) (cloud.InstanceSet, error) {
+func (sd *StubDriver) InstanceSet(params json.RawMessage, id cloud.InstanceSetID, _ cloud.SharedResourceTags, logger logrus.FieldLogger, reg *prometheus.Registry) (cloud.InstanceSet, error) {
        if sd.holdCloudOps == nil {
                sd.holdCloudOps = make(chan bool)
        }

diff --git a/lib/dispatchcloud/worker/pool_test.go b/lib/dispatchcloud/worker/pool_test.go
index 7f3a1531ea06bed7daebeef6dd913ddab11a53bd..8d2ba09ebe849f5962f7fcf1570de27f5fe3012e 100644 (file)
--- a/lib/dispatchcloud/worker/pool_test.go
+++ b/lib/dispatchcloud/worker/pool_test.go
@@ -78,7 +78,7 @@ func (suite *PoolSuite) TestResumeAfterRestart(c *check.C) {
 
        driver := &test.StubDriver{}
        instanceSetID := cloud.InstanceSetID("test-instance-set-id")
-       is, err := driver.InstanceSet(nil, instanceSetID, nil, suite.logger)
+       is, err := driver.InstanceSet(nil, instanceSetID, nil, suite.logger, nil)
        c.Assert(err, check.IsNil)
 
        newExecutor := func(cloud.Instance) Executor {
@@ -157,7 +157,7 @@ func (suite *PoolSuite) TestResumeAfterRestart(c *check.C) {
 
 func (suite *PoolSuite) TestDrain(c *check.C) {
        driver := test.StubDriver{}
-       instanceSet, err := driver.InstanceSet(nil, "test-instance-set-id", nil, suite.logger)
+       instanceSet, err := driver.InstanceSet(nil, "test-instance-set-id", nil, suite.logger, nil)
        c.Assert(err, check.IsNil)
 
        ac := arvados.NewClientFromEnv()
@@ -210,7 +210,7 @@ func (suite *PoolSuite) TestDrain(c *check.C) {
 
 func (suite *PoolSuite) TestNodeCreateThrottle(c *check.C) {
        driver := test.StubDriver{HoldCloudOps: true}
-       instanceSet, err := driver.InstanceSet(nil, "test-instance-set-id", nil, suite.logger)
+       instanceSet, err := driver.InstanceSet(nil, "test-instance-set-id", nil, suite.logger, nil)
        c.Assert(err, check.IsNil)
 
        type1 := test.InstanceType(1)
@@ -250,7 +250,7 @@ func (suite *PoolSuite) TestNodeCreateThrottle(c *check.C) {
 
 func (suite *PoolSuite) TestCreateUnallocShutdown(c *check.C) {
        driver := test.StubDriver{HoldCloudOps: true}
-       instanceSet, err := driver.InstanceSet(nil, "test-instance-set-id", nil, suite.logger)
+       instanceSet, err := driver.InstanceSet(nil, "test-instance-set-id", nil, suite.logger, nil)
        c.Assert(err, check.IsNil)
 
        type1 := arvados.InstanceType{Name: "a1s", ProviderType: "a1.small", VCPUs: 1, RAM: 1 * GiB, Price: .01}

diff --git a/lib/dispatchcloud/worker/worker_test.go b/lib/dispatchcloud/worker/worker_test.go
index d04ecbb72f3bdd8a6dff1fa942171a8ba08347cc..5d8c67e9162140a9eeb8f62a897c7c955d900188 100644 (file)
--- a/lib/dispatchcloud/worker/worker_test.go
+++ b/lib/dispatchcloud/worker/worker_test.go
@@ -43,7 +43,7 @@ func (suite *WorkerSuite) TestProbeAndUpdate(c *check.C) {
        probeTimeout := time.Second
 
        ac := arvados.NewClientFromEnv()
-       is, err := (&test.StubDriver{}).InstanceSet(nil, "test-instance-set-id", nil, suite.logger)
+       is, err := (&test.StubDriver{}).InstanceSet(nil, "test-instance-set-id", nil, suite.logger, nil)
        c.Assert(err, check.IsNil)
        inst, err := is.Create(arvados.InstanceType{}, "", nil, "echo InitCommand", nil)
        c.Assert(err, check.IsNil)