//
// SPDX-License-Identifier: AGPL-3.0
-package main
+package dispatchslurm
import (
"bytes"
"context"
"errors"
+ "flag"
"fmt"
"io"
"io/ioutil"
- "log"
"net/http"
"net/http/httptest"
"os"
"testing"
"time"
- "git.curoverse.com/arvados.git/lib/dispatchcloud"
- "git.curoverse.com/arvados.git/sdk/go/arvados"
- "git.curoverse.com/arvados.git/sdk/go/arvadosclient"
- "git.curoverse.com/arvados.git/sdk/go/arvadostest"
- "git.curoverse.com/arvados.git/sdk/go/dispatch"
+ "git.arvados.org/arvados.git/lib/cmd"
+ "git.arvados.org/arvados.git/lib/config"
+ "git.arvados.org/arvados.git/lib/dispatchcloud"
+ "git.arvados.org/arvados.git/sdk/go/arvados"
+ "git.arvados.org/arvados.git/sdk/go/arvadosclient"
+ "git.arvados.org/arvados.git/sdk/go/arvadostest"
+ "git.arvados.org/arvados.git/sdk/go/ctxlog"
+ "git.arvados.org/arvados.git/sdk/go/dispatch"
"github.com/sirupsen/logrus"
. "gopkg.in/check.v1"
)
}
func (s *IntegrationSuite) SetUpTest(c *C) {
- arvadostest.StartAPI()
+ arvadostest.ResetEnv()
+ arvadostest.ResetDB(c)
os.Setenv("ARVADOS_API_TOKEN", arvadostest.Dispatch1Token)
s.disp = Dispatcher{}
s.disp.cluster = &arvados.Cluster{}
func (s *IntegrationSuite) TearDownTest(c *C) {
arvadostest.ResetEnv()
- arvadostest.StopAPI()
+ arvadostest.ResetDB(c)
}
type slurmFake struct {
func (s *IntegrationSuite) integrationTest(c *C,
expectBatch [][]string,
- runContainer func(*dispatch.Dispatcher, arvados.Container)) arvados.Container {
+ runContainer func(*dispatch.Dispatcher, arvados.Container)) (arvados.Container, error) {
arvadostest.ResetEnv()
arv, err := arvadosclient.MakeArvadosClient()
ctx, cancel := context.WithCancel(context.Background())
doneRun := make(chan struct{})
+ doneDispatch := make(chan error)
s.disp.Dispatcher = &dispatch.Dispatcher{
Arv: arv,
PollPeriod: time.Second,
- RunContainer: func(disp *dispatch.Dispatcher, ctr arvados.Container, status <-chan arvados.Container) {
+ RunContainer: func(disp *dispatch.Dispatcher, ctr arvados.Container, status <-chan arvados.Container) error {
go func() {
runContainer(disp, ctr)
s.slurm.queue = ""
doneRun <- struct{}{}
}()
- s.disp.runContainer(disp, ctr, status)
+ err := s.disp.runContainer(disp, ctr, status)
cancel()
+ doneDispatch <- err
+ return nil
},
}
err = s.disp.Dispatcher.Run(ctx)
<-doneRun
c.Assert(err, Equals, context.Canceled)
+ errDispatch := <-doneDispatch
s.disp.sqCheck.Stop()
var container arvados.Container
err = arv.Get("containers", "zzzzz-dz642-queuedcontainer", nil, &container)
c.Check(err, IsNil)
- return container
+ return container, errDispatch
}
func (s *IntegrationSuite) TestNormal(c *C) {
s.slurm = slurmFake{queue: "zzzzz-dz642-queuedcontainer 10000 100 PENDING Resources\n"}
- container := s.integrationTest(c,
+ container, _ := s.integrationTest(c,
nil,
func(dispatcher *dispatch.Dispatcher, container arvados.Container) {
dispatcher.UpdateState(container.UUID, dispatch.Running)
s.slurm = slurmFake{queue: "zzzzz-dz642-queuedcontainer 10000 100 PENDING Resources\n"}
readyToCancel := make(chan bool)
s.slurm.onCancel = func() { <-readyToCancel }
- container := s.integrationTest(c,
+ container, _ := s.integrationTest(c,
nil,
func(dispatcher *dispatch.Dispatcher, container arvados.Container) {
dispatcher.UpdateState(container.UUID, dispatch.Running)
}
func (s *IntegrationSuite) TestMissingFromSqueue(c *C) {
- container := s.integrationTest(c,
+ container, _ := s.integrationTest(c,
[][]string{{
fmt.Sprintf("--job-name=%s", "zzzzz-dz642-queuedcontainer"),
fmt.Sprintf("--nice=%d", 10000),
func (s *IntegrationSuite) TestSbatchFail(c *C) {
s.slurm = slurmFake{errBatch: errors.New("something terrible happened")}
- container := s.integrationTest(c,
+ container, err := s.integrationTest(c,
[][]string{{"--job-name=zzzzz-dz642-queuedcontainer", "--nice=10000", "--no-requeue", "--mem=11445", "--cpus-per-task=4", "--tmp=45777"}},
func(dispatcher *dispatch.Dispatcher, container arvados.Container) {
dispatcher.UpdateState(container.UUID, dispatch.Running)
dispatcher.UpdateState(container.UUID, dispatch.Complete)
})
c.Check(container.State, Equals, arvados.ContainerStateComplete)
-
- arv, err := arvadosclient.MakeArvadosClient()
- c.Assert(err, IsNil)
-
- var ll arvados.LogList
- err = arv.List("logs", arvadosclient.Dict{"filters": [][]string{
- {"object_uuid", "=", container.UUID},
- {"event_type", "=", "dispatch"},
- }}, &ll)
- c.Assert(err, IsNil)
- c.Assert(len(ll.Items), Equals, 1)
+ c.Check(err, ErrorMatches, `something terrible happened`)
}
type StubbedSuite struct {
dispatcher := dispatch.Dispatcher{
Arv: arv,
PollPeriod: time.Second,
- RunContainer: func(disp *dispatch.Dispatcher, ctr arvados.Container, status <-chan arvados.Container) {
+ RunContainer: func(disp *dispatch.Dispatcher, ctr arvados.Container, status <-chan arvados.Container) error {
go func() {
time.Sleep(time.Second)
disp.UpdateState(ctr.UUID, dispatch.Running)
}()
s.disp.runContainer(disp, ctr, status)
cancel()
+ return nil
},
}
}
func (s *StubbedSuite) TestLoadLegacyConfig(c *C) {
+ log := ctxlog.TestLogger(c)
content := []byte(`
Client:
APIHost: example.com
AuthToken: abcdefg
+ KeepServiceURIs:
+ - https://example.com/keep1
+ - https://example.com/keep2
SbatchArguments: ["--foo", "bar"]
PollPeriod: 12s
PrioritySpread: 42
MinRetryPeriod: 13s
BatchSize: 99
`)
- tmpfile, err := ioutil.TempFile("", "example")
- if err != nil {
- log.Fatal(err)
- }
+ tmpfile := c.MkDir() + "/config.yml"
+ err := ioutil.WriteFile(tmpfile, content, 0777)
+ c.Assert(err, IsNil)
- defer os.Remove(tmpfile.Name()) // clean up
+ os.Setenv("ARVADOS_KEEP_SERVICES", "")
- if _, err := tmpfile.Write(content); err != nil {
- log.Fatal(err)
- }
- if err := tmpfile.Close(); err != nil {
- log.Fatal(err)
+ flags := flag.NewFlagSet("", flag.ContinueOnError)
+ flags.SetOutput(os.Stderr)
+ loader := config.NewLoader(&bytes.Buffer{}, log)
+ loader.SetupFlags(flags)
+ args := loader.MungeLegacyConfigArgs(log, []string{"-config", tmpfile}, "-legacy-"+string(arvados.ServiceNameDispatchSLURM)+"-config")
+ ok, _ := cmd.ParseFlags(flags, "crunch-dispatch-slurm", args, "", os.Stderr)
+ c.Check(ok, Equals, true)
+ cfg, err := loader.Load()
+ c.Assert(err, IsNil)
+ cluster, err := cfg.GetCluster("")
+ c.Assert(err, IsNil)
- }
- err = s.disp.configure("crunch-dispatch-slurm", []string{"-config", tmpfile.Name()})
- c.Check(err, IsNil)
+ c.Check(cluster.Services.Controller.ExternalURL, Equals, arvados.URL{Scheme: "https", Host: "example.com", Path: "/"})
+ c.Check(cluster.SystemRootToken, Equals, "abcdefg")
+ c.Check(cluster.Containers.SLURM.SbatchArgumentsList, DeepEquals, []string{"--foo", "bar"})
+ c.Check(cluster.Containers.CloudVMs.PollInterval, Equals, arvados.Duration(12*time.Second))
+ c.Check(cluster.Containers.SLURM.PrioritySpread, Equals, int64(42))
+ c.Check(cluster.Containers.CrunchRunCommand, Equals, "x-crunch-run")
+ c.Check(cluster.Containers.CrunchRunArgumentsList, DeepEquals, []string{"--cgroup-parent-subsystem=memory"})
+ c.Check(cluster.Containers.ReserveExtraRAM, Equals, arvados.ByteSize(12345))
+ c.Check(cluster.Containers.MinRetryPeriod, Equals, arvados.Duration(13*time.Second))
+ c.Check(cluster.API.MaxItemsPerResponse, Equals, 99)
+ c.Check(cluster.Containers.SLURM.SbatchEnvironmentVariables, DeepEquals, map[string]string{
+ "ARVADOS_KEEP_SERVICES": "https://example.com/keep1 https://example.com/keep2",
+ })
- c.Check(s.disp.cluster.Services.Controller.ExternalURL, Equals, arvados.URL{Scheme: "https", Host: "example.com"})
- c.Check(s.disp.cluster.SystemRootToken, Equals, "abcdefg")
- c.Check(s.disp.cluster.Containers.SLURM.SbatchArgumentsList, DeepEquals, []string{"--foo", "bar"})
- c.Check(s.disp.cluster.Containers.CloudVMs.PollInterval, Equals, arvados.Duration(12*time.Second))
- c.Check(s.disp.cluster.Containers.SLURM.PrioritySpread, Equals, int64(42))
- c.Check(s.disp.cluster.Containers.CrunchRunCommand, Equals, "x-crunch-run")
- c.Check(s.disp.cluster.Containers.CrunchRunArgumentsList, DeepEquals, []string{"--cgroup-parent-subsystem=memory"})
- c.Check(s.disp.cluster.Containers.ReserveExtraRAM, Equals, arvados.ByteSize(12345))
- c.Check(s.disp.cluster.Containers.MinRetryPeriod, Equals, arvados.Duration(13*time.Second))
- c.Check(s.disp.cluster.API.MaxItemsPerResponse, Equals, 99)
+ // Ensure configure() copies SbatchEnvironmentVariables into
+ // the current process's environment (that's how they end up
+ // getting passed to sbatch).
+ s.disp.cluster = cluster
+ s.disp.configure()
+ c.Check(os.Getenv("ARVADOS_KEEP_SERVICES"), Equals, "https://example.com/keep1 https://example.com/keep2")
}