Merge pull request #1 from arvados/master

[arvados.git] / lib / dispatchcloud / worker / pool_test.go

diff --git a/lib/dispatchcloud/worker/pool_test.go b/lib/dispatchcloud/worker/pool_test.go
index 526bc552cd7dc56dee698b026d7479ed4b0840a8..1948c1e874859f2d8355115b3671f2c5ef0ae32d 100644 (file)
--- a/lib/dispatchcloud/worker/pool_test.go
+++ b/lib/dispatchcloud/worker/pool_test.go
@@ -9,9 +9,10 @@ import (
        "strings"
        "time"
 
-       "git.curoverse.com/arvados.git/lib/cloud"
-       "git.curoverse.com/arvados.git/lib/dispatchcloud/test"
-       "git.curoverse.com/arvados.git/sdk/go/arvados"
+       "git.arvados.org/arvados.git/lib/cloud"
+       "git.arvados.org/arvados.git/lib/dispatchcloud/test"
+       "git.arvados.org/arvados.git/sdk/go/arvados"
+       "git.arvados.org/arvados.git/sdk/go/ctxlog"
        "github.com/prometheus/client_golang/prometheus"
        check "gopkg.in/check.v1"
 )
@@ -62,26 +63,30 @@ func (suite *PoolSuite) TestResumeAfterRestart(c *check.C) {
                }
        }
 
-       logger := test.Logger()
+       logger := ctxlog.TestLogger(c)
        driver := &test.StubDriver{}
-       is, err := driver.InstanceSet(nil, "", logger)
+       instanceSetID := cloud.InstanceSetID("test-instance-set-id")
+       is, err := driver.InstanceSet(nil, instanceSetID, nil, logger)
        c.Assert(err, check.IsNil)
 
        newExecutor := func(cloud.Instance) Executor {
-               return stubExecutor{
-                       "crunch-run --list": stubResp{},
-                       "true":              stubResp{},
+               return &stubExecutor{
+                       response: map[string]stubResp{
+                               "crunch-run --list": stubResp{},
+                               "true":              stubResp{},
+                       },
                }
        }
 
        cluster := &arvados.Cluster{
-               Dispatch: arvados.Dispatch{
-                       MaxProbesPerSecond: 1000,
-                       ProbeInterval:      arvados.Duration(time.Millisecond * 10),
-               },
-               CloudVMs: arvados.CloudVMs{
-                       BootProbeCommand: "true",
-                       SyncInterval:     arvados.Duration(time.Millisecond * 10),
+               Containers: arvados.ContainersConfig{
+                       CloudVMs: arvados.CloudVMsConfig{
+                               BootProbeCommand:   "true",
+                               MaxProbesPerSecond: 1000,
+                               ProbeInterval:      arvados.Duration(time.Millisecond * 10),
+                               SyncInterval:       arvados.Duration(time.Millisecond * 10),
+                               TagKeyPrefix:       "testprefix:",
+                       },
                },
                InstanceTypes: arvados.InstanceTypeMap{
                        type1.Name: type1,
@@ -90,7 +95,7 @@ func (suite *PoolSuite) TestResumeAfterRestart(c *check.C) {
                },
        }
 
-       pool := NewPool(logger, arvados.NewClientFromEnv(), prometheus.NewRegistry(), is, newExecutor, nil, cluster)
+       pool := NewPool(logger, arvados.NewClientFromEnv(), prometheus.NewRegistry(), instanceSetID, is, newExecutor, nil, cluster)
        notify := pool.Subscribe()
        defer pool.Unsubscribe(notify)
        pool.Create(type1)
@@ -104,11 +109,29 @@ func (suite *PoolSuite) TestResumeAfterRestart(c *check.C) {
                        pool.SetIdleBehavior(heldInstanceID, IdleBehaviorHold)
                }
        }
+       // Wait for the tags to save to the cloud provider
+       tagKey := cluster.Containers.CloudVMs.TagKeyPrefix + tagKeyIdleBehavior
+       deadline := time.Now().Add(time.Second)
+       for !func() bool {
+               pool.mtx.RLock()
+               defer pool.mtx.RUnlock()
+               for _, wkr := range pool.workers {
+                       if wkr.instType == type2 {
+                               return wkr.instance.Tags()[tagKey] == string(IdleBehaviorHold)
+                       }
+               }
+               return false
+       }() {
+               if time.Now().After(deadline) {
+                       c.Fatal("timeout")
+               }
+               time.Sleep(time.Millisecond * 10)
+       }
        pool.Stop()
 
        c.Log("------- starting new pool, waiting to recover state")
 
-       pool2 := NewPool(logger, arvados.NewClientFromEnv(), prometheus.NewRegistry(), is, newExecutor, nil, cluster)
+       pool2 := NewPool(logger, arvados.NewClientFromEnv(), prometheus.NewRegistry(), instanceSetID, is, newExecutor, nil, cluster)
        notify2 := pool2.Subscribe()
        defer pool2.Unsubscribe(notify2)
        waitForIdle(pool2, notify2)
@@ -124,9 +147,9 @@ func (suite *PoolSuite) TestResumeAfterRestart(c *check.C) {
 }
 
 func (suite *PoolSuite) TestCreateUnallocShutdown(c *check.C) {
-       logger := test.Logger()
+       logger := ctxlog.TestLogger(c)
        driver := test.StubDriver{HoldCloudOps: true}
-       instanceSet, err := driver.InstanceSet(nil, "", logger)
+       instanceSet, err := driver.InstanceSet(nil, "test-instance-set-id", nil, logger)
        c.Assert(err, check.IsNil)
 
        type1 := arvados.InstanceType{Name: "a1s", ProviderType: "a1.small", VCPUs: 1, RAM: 1 * GiB, Price: .01}
@@ -134,7 +157,7 @@ func (suite *PoolSuite) TestCreateUnallocShutdown(c *check.C) {
        type3 := arvados.InstanceType{Name: "a2l", ProviderType: "a2.large", VCPUs: 4, RAM: 4 * GiB, Price: .04}
        pool := &Pool{
                logger:      logger,
-               newExecutor: func(cloud.Instance) Executor { return stubExecutor{} },
+               newExecutor: func(cloud.Instance) Executor { return &stubExecutor{} },
                instanceSet: &throttledInstanceSet{InstanceSet: instanceSet},
                instanceTypes: arvados.InstanceTypeMap{
                        type1.Name: type1,