-package main
+// Copyright (C) The Arvados Authors. All rights reserved.
+//
+// SPDX-License-Identifier: AGPL-3.0
+
+package keepstore
import (
"bytes"
+ "context"
+ "crypto/md5"
+ "encoding/json"
"errors"
- "git.curoverse.com/arvados.git/sdk/go/arvadosclient"
- "git.curoverse.com/arvados.git/sdk/go/keepclient"
- . "gopkg.in/check.v1"
+ "fmt"
"io"
"net/http"
- "strings"
- "testing"
+ "net/http/httptest"
+ "sort"
"time"
-)
-
-var testPullLists map[string]string
-var processedPullLists map[string]string
-
-type PullWorkerTestSuite struct{}
-
-// Gocheck boilerplate
-func TestPullWorker(t *testing.T) {
- TestingT(t)
-}
-
-// Gocheck boilerplate
-var _ = Suite(&PullWorkerTestSuite{})
-
-func (s *PullWorkerTestSuite) SetUpSuite(c *C) {
- // Since keepstore does not come into picture in tests,
- // we need to explicitly start the goroutine in tests.
- arv, err := arvadosclient.MakeArvadosClient()
- c.Assert(err, Equals, nil)
- keepClient, err := keepclient.MakeKeepClient(&arv)
- c.Assert(err, Equals, nil)
- go RunPullWorker(pullq, keepClient)
-
- // When a new pull request arrives, the old one will be overwritten.
- // This behavior is simulated with delay tests below.
- testPullLists = make(map[string]string)
- processedPullLists = make(map[string]string)
-}
-
-func (s *PullWorkerTestSuite) TearDownSuite(c *C) {
- // give the channel some time to read and process all pull list entries
- time.Sleep(1000 * time.Millisecond)
-
- expectWorkerChannelEmpty(c, pullq.NextItem)
-
- c.Assert(len(processedPullLists), Not(Equals), len(testPullLists))
-}
-
-var first_pull_list = []byte(`[
- {
- "locator":"locator1",
- "servers":[
- "server_1",
- "server_2"
- ]
- },
- {
- "locator":"locator2",
- "servers":[
- "server_3"
- ]
- }
- ]`)
-
-var second_pull_list = []byte(`[
- {
- "locator":"locator3",
- "servers":[
- "server_1",
- "server_2"
- ]
- }
- ]`)
-
-type PullWorkerTestData struct {
- name string
- req RequestTester
- response_code int
- response_body string
- read_content string
- read_error bool
- put_error bool
-}
-
-func (s *PullWorkerTestSuite) TestPullWorker_pull_list_with_two_locators(c *C) {
- defer teardown()
-
- data_manager_token = "DATA MANAGER TOKEN"
- testData := PullWorkerTestData{
- "TestPullWorker_pull_list_with_two_locators",
- RequestTester{"/pull", data_manager_token, "PUT", first_pull_list},
- http.StatusOK,
- "Received 2 pull requests\n",
- "hello",
- false,
- false,
- }
-
- performTest(testData, c)
-}
-
-func (s *PullWorkerTestSuite) TestPullWorker_pull_list_with_one_locator(c *C) {
- defer teardown()
-
- data_manager_token = "DATA MANAGER TOKEN"
- testData := PullWorkerTestData{
- "TestPullWorker_pull_list_with_one_locator",
- RequestTester{"/pull", data_manager_token, "PUT", second_pull_list},
- http.StatusOK,
- "Received 1 pull requests\n",
- "hola",
- false,
- false,
- }
-
- performTest(testData, c)
-}
-
-// When a new pull request arrives, the old one will be overwritten.
-// Simulate this behavior by inducing delay in GetContent for the delay test(s).
-// To ensure this delay test is not the last one executed and
-// hence we cannot verify this behavior, let's run the delay test twice.
-func (s *PullWorkerTestSuite) TestPullWorker_pull_list_with_one_locator_with_delay_1(c *C) {
- defer teardown()
-
- data_manager_token = "DATA MANAGER TOKEN"
- testData := PullWorkerTestData{
- "TestPullWorker_pull_list_with_one_locator_with_delay_1",
- RequestTester{"/pull", data_manager_token, "PUT", second_pull_list},
- http.StatusOK,
- "Received 1 pull requests\n",
- "hola",
- false,
- false,
- }
-
- performTest(testData, c)
-}
-func (s *PullWorkerTestSuite) TestPullWorker_pull_list_with_one_locator_with_delay_2(c *C) {
- defer teardown()
-
- data_manager_token = "DATA MANAGER TOKEN"
- testData := PullWorkerTestData{
- "TestPullWorker_pull_list_with_one_locator_with_delay_2",
- RequestTester{"/pull", data_manager_token, "PUT", second_pull_list},
- http.StatusOK,
- "Received 1 pull requests\n",
- "hola",
- false,
- false,
- }
-
- performTest(testData, c)
-}
-
-func (s *PullWorkerTestSuite) TestPullWorker_error_on_get_one_locator(c *C) {
- defer teardown()
-
- data_manager_token = "DATA MANAGER TOKEN"
- testData := PullWorkerTestData{
- "TestPullWorker_error_on_get_one_locator",
- RequestTester{"/pull", data_manager_token, "PUT", second_pull_list},
- http.StatusOK,
- "Received 1 pull requests\n",
- "unused",
- true,
- false,
- }
-
- performTest(testData, c)
-}
-
-func (s *PullWorkerTestSuite) TestPullWorker_error_on_get_two_locators(c *C) {
- defer teardown()
-
- data_manager_token = "DATA MANAGER TOKEN"
- testData := PullWorkerTestData{
- "TestPullWorker_error_on_get_two_locators",
- RequestTester{"/pull", data_manager_token, "PUT", first_pull_list},
- http.StatusOK,
- "Received 2 pull requests\n",
- "unused",
- true,
- false,
- }
-
- performTest(testData, c)
-}
-
-func (s *PullWorkerTestSuite) TestPullWorker_error_on_put_one_locator(c *C) {
- defer teardown()
-
- data_manager_token = "DATA MANAGER TOKEN"
- testData := PullWorkerTestData{
- "TestPullWorker_error_on_put_one_locator",
- RequestTester{"/pull", data_manager_token, "PUT", second_pull_list},
- http.StatusOK,
- "Received 1 pull requests\n",
- "unused",
- false,
- true,
- }
-
- performTest(testData, c)
-}
-
-func (s *PullWorkerTestSuite) TestPullWorker_error_on_put_two_locators(c *C) {
- defer teardown()
+ "git.arvados.org/arvados.git/sdk/go/arvados"
+ "git.arvados.org/arvados.git/sdk/go/arvadostest"
+ "github.com/sirupsen/logrus"
+ . "gopkg.in/check.v1"
+)
- data_manager_token = "DATA MANAGER TOKEN"
- testData := PullWorkerTestData{
- "TestPullWorker_error_on_put_two_locators",
- RequestTester{"/pull", data_manager_token, "PUT", first_pull_list},
- http.StatusOK,
- "Received 2 pull requests\n",
- "unused",
- false,
- true,
+func (s *routerSuite) TestPullList_Execute(c *C) {
+ remotecluster := testCluster(c)
+ remotecluster.Volumes = map[string]arvados.Volume{
+ "zzzzz-nyw5e-rrrrrrrrrrrrrrr": {Replication: 1, Driver: "stub"},
}
-
- performTest(testData, c)
-}
-
-func performTest(testData PullWorkerTestData, c *C) {
- testPullLists[testData.name] = testData.response_body
-
- // We need to make sure the tests have a slight delay so that we can verify the pull list channel overwrites.
- time.Sleep(25 * time.Millisecond)
-
- // Override GetContent to mock keepclient Get functionality
- GetContent = func(signedLocator string, keepClient keepclient.KeepClient) (
- reader io.ReadCloser, contentLength int64, url string, err error) {
- if strings.HasPrefix(testData.name, "TestPullWorker_pull_list_with_one_locator_with_delay") {
- time.Sleep(100 * time.Millisecond)
- }
-
- processedPullLists[testData.name] = testData.response_body
- if testData.read_error {
- return nil, 0, "", errors.New("Error getting data")
- } else {
- cb := &ClosingBuffer{bytes.NewBufferString("Hi!")}
- var rc io.ReadCloser
- rc = cb
- return rc, 3, "", nil
+ remoterouter, cancel := testRouter(c, remotecluster, nil)
+ defer cancel()
+ remoteserver := httptest.NewServer(remoterouter)
+ defer remoteserver.Close()
+
+ router, cancel := testRouter(c, s.cluster, nil)
+ defer cancel()
+
+ executePullList := func(pullList []PullListItem) string {
+ var logbuf bytes.Buffer
+ logger := logrus.New()
+ logger.Out = &logbuf
+ router.keepstore.logger = logger
+
+ listjson, err := json.Marshal(pullList)
+ c.Assert(err, IsNil)
+ resp := call(router, "PUT", "http://example/pull", s.cluster.SystemRootToken, listjson, nil)
+ c.Check(resp.Code, Equals, http.StatusOK)
+ for {
+ router.puller.cond.L.Lock()
+ todolen := len(router.puller.todo)
+ router.puller.cond.L.Unlock()
+ if todolen == 0 && router.puller.inprogress.Load() == 0 {
+ break
+ }
+ time.Sleep(time.Millisecond)
}
+ return logbuf.String()
}
- // Override PutContent to mock PutBlock functionality
- PutContent = func(content []byte, locator string) (err error) {
- if testData.put_error {
- return errors.New("Error putting data")
- } else {
- return nil
- }
+ newRemoteBlock := func(datastring string) string {
+ data := []byte(datastring)
+ hash := fmt.Sprintf("%x", md5.Sum(data))
+ locator := fmt.Sprintf("%s+%d", hash, len(data))
+ _, err := remoterouter.keepstore.BlockWrite(context.Background(), arvados.BlockWriteOptions{
+ Hash: hash,
+ Data: data,
+ })
+ c.Assert(err, IsNil)
+ return locator
}
- response := IssueRequest(&testData.req)
- c.Assert(testData.response_code, Equals, response.Code)
- c.Assert(testData.response_body, Equals, response.Body.String())
-}
+ mounts := append([]*mount(nil), router.keepstore.mountsR...)
+ sort.Slice(mounts, func(i, j int) bool { return mounts[i].UUID < mounts[j].UUID })
+ var vols []*stubVolume
+ for _, mount := range mounts {
+ vols = append(vols, mount.volume.(*stubVolume))
+ }
-type ClosingBuffer struct {
- *bytes.Buffer
-}
+ ctx := authContext(arvadostest.ActiveTokenV2)
-func (cb *ClosingBuffer) Close() (err error) {
- return
-}
+ locator := newRemoteBlock("pull available block to unspecified volume")
+ executePullList([]PullListItem{{
+ Locator: locator,
+ Servers: []string{remoteserver.URL}}})
+ _, err := router.keepstore.BlockRead(ctx, arvados.BlockReadOptions{
+ Locator: router.keepstore.signLocator(arvadostest.ActiveTokenV2, locator),
+ WriteTo: io.Discard})
+ c.Check(err, IsNil)
-func expectWorkerChannelEmpty(c *C, workerChannel <-chan interface{}) {
- select {
- case item := <-workerChannel:
- c.Fatalf("Received value (%v) from channel that was expected to be empty", item)
- default:
- }
+ locator0 := newRemoteBlock("pull available block to specified volume 0")
+ locator1 := newRemoteBlock("pull available block to specified volume 1")
+ executePullList([]PullListItem{
+ {
+ Locator: locator0,
+ Servers: []string{remoteserver.URL},
+ MountUUID: vols[0].params.UUID},
+ {
+ Locator: locator1,
+ Servers: []string{remoteserver.URL},
+ MountUUID: vols[1].params.UUID}})
+ c.Check(vols[0].data[locator0[:32]].data, NotNil)
+ c.Check(vols[1].data[locator1[:32]].data, NotNil)
+
+ locator = fooHash + "+3"
+ logs := executePullList([]PullListItem{{
+ Locator: locator,
+ Servers: []string{remoteserver.URL}}})
+ c.Check(logs, Matches, ".*error pulling data from remote servers.*Block not found.*locator=acbd.*\n")
+
+ locator = fooHash + "+3"
+ logs = executePullList([]PullListItem{{
+ Locator: locator,
+ Servers: []string{"http://0.0.0.0:9/"}}})
+ c.Check(logs, Matches, ".*error pulling data from remote servers.*connection refused.*locator=acbd.*\n")
+
+ locator = newRemoteBlock("log error writing to local volume")
+ vols[0].blockWrite = func(context.Context, string, []byte) error { return errors.New("test error") }
+ vols[1].blockWrite = vols[0].blockWrite
+ logs = executePullList([]PullListItem{{
+ Locator: locator,
+ Servers: []string{remoteserver.URL}}})
+ c.Check(logs, Matches, ".*error writing data to zzzzz-nyw5e-.*error=\"test error\".*locator=.*\n")
+ vols[0].blockWrite = nil
+ vols[1].blockWrite = nil
+
+ locator = newRemoteBlock("log error when destination mount does not exist")
+ logs = executePullList([]PullListItem{{
+ Locator: locator,
+ Servers: []string{remoteserver.URL},
+ MountUUID: "bogus-mount-uuid"}})
+ c.Check(logs, Matches, ".*ignoring pull list entry for nonexistent mount bogus-mount-uuid.*locator=.*\n")
+
+ logs = executePullList([]PullListItem{})
+ c.Logf("%s", logs)
}