X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/8678b8542a1e73c9a7861a9b8ba7488a1a32c1c4..3272913e4a91f5b80b256f9f09e4d4f432d07058:/services/datamanager/datamanager_test.go diff --git a/services/datamanager/datamanager_test.go b/services/datamanager/datamanager_test.go index c76d481f23..094cd44b0f 100644 --- a/services/datamanager/datamanager_test.go +++ b/services/datamanager/datamanager_test.go @@ -16,11 +16,6 @@ import ( "time" ) -const ( - ActiveUserToken = "3kg6k6lzmp9kj5cpkcoxie963cmvjahbt2fod9zru30k1jqdmi" - AdminToken = "4axaw8zxe0qm22wa6urpp5nskcne8z88cvbupv653y1njyi05h" -) - var arv arvadosclient.ArvadosClient var keepClient *keepclient.KeepClient var keepServers []string @@ -31,9 +26,10 @@ func SetupDataManagerTest(t *testing.T) { // start api and keep servers arvadostest.ResetEnv() arvadostest.StartAPI() - arvadostest.StartKeep() + arvadostest.StartKeep(2, false) arv = makeArvadosClient() + arv.ApiToken = arvadostest.DataManagerToken // keep client keepClient = &keepclient.KeepClient{ @@ -47,13 +43,14 @@ func SetupDataManagerTest(t *testing.T) { if err := keepClient.DiscoverKeepServers(); err != nil { t.Fatalf("Error discovering keep services: %s", err) } + keepServers = []string{} for _, host := range keepClient.LocalRoots() { keepServers = append(keepServers, host) } } func TearDownDataManagerTest(t *testing.T) { - arvadostest.StopKeep() + arvadostest.StopKeep(2) arvadostest.StopAPI() } @@ -123,7 +120,18 @@ func getFirstLocatorFromCollection(t *testing.T, uuid string) string { return match[1] + "+" + match[2] } +func switchToken(t string) func() { + orig := arv.ApiToken + restore := func() { + arv.ApiToken = orig + } + arv.ApiToken = t + return restore +} + func getCollection(t *testing.T, uuid string) Dict { + defer switchToken(arvadostest.AdminToken)() + getback := make(Dict) err := arv.Get("collections", uuid, nil, &getback) if err != nil { @@ -137,6 +145,8 @@ func getCollection(t *testing.T, uuid string) Dict { } func updateCollection(t *testing.T, uuid string, paramName string, paramValue string) { + defer switchToken(arvadostest.AdminToken)() + err := arv.Update("collections", uuid, arvadosclient.Dict{ "collection": arvadosclient.Dict{ paramName: paramValue, @@ -151,6 +161,8 @@ func updateCollection(t *testing.T, uuid string, paramName string, paramValue st type Dict map[string]interface{} func deleteCollection(t *testing.T, uuid string) { + defer switchToken(arvadostest.AdminToken)() + getback := make(Dict) err := arv.Delete("collections", uuid, nil, &getback) if err != nil { @@ -174,7 +186,7 @@ func getBlockIndexesForServer(t *testing.T, i int) []string { path := keepServers[i] + "/index" client := http.Client{} req, err := http.NewRequest("GET", path, nil) - req.Header.Add("Authorization", "OAuth2 " + AdminToken) + req.Header.Add("Authorization", "OAuth2 "+arvadostest.DataManagerToken) req.Header.Add("Content-Type", "application/octet-stream") resp, err := client.Do(req) defer resp.Body.Close() @@ -205,7 +217,7 @@ func getBlockIndexes(t *testing.T) [][]string { return indexes } -func verifyBlocks(t *testing.T, notExpected []string, expected []string) { +func verifyBlocks(t *testing.T, notExpected []string, expected []string, minReplication int) { blocks := getBlockIndexes(t) for _, block := range notExpected { @@ -223,8 +235,8 @@ func verifyBlocks(t *testing.T, notExpected []string, expected []string) { nFound++ } } - if nFound < 2 { - t.Fatalf("Found %d replicas of block %s, expected >= 2", nFound, block) + if nFound < minReplication { + t.Fatalf("Found %d replicas of block %s, expected >= %d", nFound, block, minReplication) } } } @@ -296,14 +308,13 @@ func backdateBlocks(t *testing.T, oldUnusedBlockLocators []string) { func getStatus(t *testing.T, path string) interface{} { client := http.Client{} req, err := http.NewRequest("GET", path, nil) - req.Header.Add("Authorization", "OAuth2 " + AdminToken) + req.Header.Add("Authorization", "OAuth2 "+arvadostest.DataManagerToken) req.Header.Add("Content-Type", "application/octet-stream") resp, err := client.Do(req) - defer resp.Body.Close() - if err != nil { t.Fatalf("Error during %s %s", path, err) } + defer resp.Body.Close() var s interface{} json.NewDecoder(resp.Body).Decode(&s) @@ -311,28 +322,18 @@ func getStatus(t *testing.T, path string) interface{} { return s } +// Wait until PullQueue and TrashQueue are empty on all keepServers. func waitUntilQueuesFinishWork(t *testing.T) { - // Wait until PullQueue and TrashQueue finish their work - for { - var done [2]bool - for i := 0; i < 2; i++ { - s := getStatus(t, keepServers[i]+"/status.json") - var pullQueueStatus interface{} - pullQueueStatus = s.(map[string]interface{})["PullQueue"] - var trashQueueStatus interface{} - trashQueueStatus = s.(map[string]interface{})["TrashQueue"] - - if pullQueueStatus.(map[string]interface{})["Queued"] == float64(0) && - pullQueueStatus.(map[string]interface{})["InProgress"] == float64(0) && - trashQueueStatus.(map[string]interface{})["Queued"] == float64(0) && - trashQueueStatus.(map[string]interface{})["InProgress"] == float64(0) { - done[i] = true - } - } - if done[0] && done[1] { - break - } else { + for _, ks := range keepServers { + for done := false; !done; { time.Sleep(100 * time.Millisecond) + s := getStatus(t, ks+"/status.json") + for _, qName := range []string{"PullQueue", "TrashQueue"} { + qStatus := s.(map[string]interface{})[qName].(map[string]interface{}) + if qStatus["Queued"].(float64)+qStatus["InProgress"].(float64) == 0 { + done = true + } + } } } } @@ -375,30 +376,34 @@ func TestPutAndGetBlocks(t *testing.T) { } // Create a collection that would be deleted later on - toBeDeletedCollectionUuid := createCollection(t, "some data for collection creation") - toBeDeletedCollectionLocator := getFirstLocatorFromCollection(t, toBeDeletedCollectionUuid) + toBeDeletedCollectionUUID := createCollection(t, "some data for collection creation") + toBeDeletedCollectionLocator := getFirstLocatorFromCollection(t, toBeDeletedCollectionUUID) // Create another collection that has the same data as the one of the old blocks - oldUsedBlockCollectionUuid := createCollection(t, oldUsedBlockData) - oldUsedBlockCollectionLocator := getFirstLocatorFromCollection(t, oldUsedBlockCollectionUuid) + oldUsedBlockCollectionUUID := createCollection(t, oldUsedBlockData) + oldUsedBlockCollectionLocator := getFirstLocatorFromCollection(t, oldUsedBlockCollectionUUID) if oldUsedBlockCollectionLocator != oldUsedBlockLocator { t.Fatalf("Locator of the collection with the same data as old block is different %s", oldUsedBlockCollectionLocator) } // Create another collection whose replication level will be changed - replicationCollectionUuid := createCollection(t, "replication level on this collection will be reduced") - replicationCollectionLocator := getFirstLocatorFromCollection(t, replicationCollectionUuid) + replicationCollectionUUID := createCollection(t, "replication level on this collection will be reduced") + replicationCollectionLocator := getFirstLocatorFromCollection(t, replicationCollectionUUID) // Create two collections with same data; one will be deleted later on dataForTwoCollections := "one of these collections will be deleted" - oneOfTwoWithSameDataUuid := createCollection(t, dataForTwoCollections) - oneOfTwoWithSameDataLocator := getFirstLocatorFromCollection(t, oneOfTwoWithSameDataUuid) - secondOfTwoWithSameDataUuid := createCollection(t, dataForTwoCollections) - secondOfTwoWithSameDataLocator := getFirstLocatorFromCollection(t, secondOfTwoWithSameDataUuid) + oneOfTwoWithSameDataUUID := createCollection(t, dataForTwoCollections) + oneOfTwoWithSameDataLocator := getFirstLocatorFromCollection(t, oneOfTwoWithSameDataUUID) + secondOfTwoWithSameDataUUID := createCollection(t, dataForTwoCollections) + secondOfTwoWithSameDataLocator := getFirstLocatorFromCollection(t, secondOfTwoWithSameDataUUID) if oneOfTwoWithSameDataLocator != secondOfTwoWithSameDataLocator { t.Fatalf("Locators for both these collections expected to be same: %s %s", oneOfTwoWithSameDataLocator, secondOfTwoWithSameDataLocator) } + // create collection with empty manifest text + emptyBlockLocator := putBlock(t, "") + emptyCollection := createCollection(t, "") + // Verify blocks before doing any backdating / deleting. var expected []string expected = append(expected, oldUnusedBlockLocators...) @@ -407,19 +412,22 @@ func TestPutAndGetBlocks(t *testing.T) { expected = append(expected, replicationCollectionLocator) expected = append(expected, oneOfTwoWithSameDataLocator) expected = append(expected, secondOfTwoWithSameDataLocator) + expected = append(expected, emptyBlockLocator) - verifyBlocks(t, nil, expected) + verifyBlocks(t, nil, expected, 2) // Run datamanager in singlerun mode dataManagerSingleRun(t) waitUntilQueuesFinishWork(t) - verifyBlocks(t, nil, expected) + verifyBlocks(t, nil, expected, 2) // Backdate the to-be old blocks and delete the collections backdateBlocks(t, oldUnusedBlockLocators) - deleteCollection(t, toBeDeletedCollectionUuid) - deleteCollection(t, secondOfTwoWithSameDataUuid) + deleteCollection(t, toBeDeletedCollectionUUID) + deleteCollection(t, secondOfTwoWithSameDataUUID) + backdateBlocks(t, []string{emptyBlockLocator}) + deleteCollection(t, emptyCollection) // Run data manager again dataManagerSingleRun(t) @@ -430,44 +438,36 @@ func TestPutAndGetBlocks(t *testing.T) { expected = append(expected, oldUsedBlockLocator) expected = append(expected, newBlockLocators...) expected = append(expected, toBeDeletedCollectionLocator) - expected = append(expected, replicationCollectionLocator) expected = append(expected, oneOfTwoWithSameDataLocator) expected = append(expected, secondOfTwoWithSameDataLocator) + expected = append(expected, emptyBlockLocator) // even when unreferenced, this remains - verifyBlocks(t, oldUnusedBlockLocators, expected) - - // Reduce replication on replicationCollectionUuid collection and verify that the overreplicated blocks are untouched. + verifyBlocks(t, oldUnusedBlockLocators, expected, 2) - // Default replication level is 2; first verify that the replicationCollectionLocator appears in both volumes - for i := 0; i < len(keepServers); i++ { - indexes := getBlockIndexesForServer(t, i) - if !valueInArray(replicationCollectionLocator, indexes) { - t.Fatalf("Not found block in index %s", replicationCollectionLocator) - } - } + // Reduce desired replication on replicationCollectionUUID + // collection, and verify that Data Manager does not reduce + // actual replication any further than that. (It might not + // reduce actual replication at all; that's OK for this test.) - // Now reduce replication level on this collection and verify that it still appears in both volumes - updateCollection(t, replicationCollectionUuid, "replication_desired", "1") - collection := getCollection(t, replicationCollectionUuid) + // Reduce desired replication level. + updateCollection(t, replicationCollectionUUID, "replication_desired", "1") + collection := getCollection(t, replicationCollectionUUID) if collection["replication_desired"].(interface{}) != float64(1) { t.Fatalf("After update replication_desired is not 1; instead it is %v", collection["replication_desired"]) } + // Verify data is currently overreplicated. + verifyBlocks(t, nil, []string{replicationCollectionLocator}, 2) + // Run data manager again - time.Sleep(100 * time.Millisecond) dataManagerSingleRun(t) waitUntilQueuesFinishWork(t) - for i := 0; i < len(keepServers); i++ { - indexes := getBlockIndexesForServer(t, i) - if !valueInArray(replicationCollectionLocator, indexes) { - t.Fatalf("Not found block in index %s", replicationCollectionLocator) - } - } - // Done testing reduce replication on collection + // Verify data is not underreplicated. + verifyBlocks(t, nil, []string{replicationCollectionLocator}, 1) - // Verify blocks one more time - verifyBlocks(t, oldUnusedBlockLocators, expected) + // Verify *other* collections' data is not underreplicated. + verifyBlocks(t, oldUnusedBlockLocators, expected, 2) } func TestDatamanagerSingleRunRepeatedly(t *testing.T) { @@ -479,13 +479,10 @@ func TestDatamanagerSingleRunRepeatedly(t *testing.T) { if err != nil { t.Fatalf("Got an error during datamanager singlerun: %v", err) } - time.Sleep(100 * time.Millisecond) } } func TestGetStatusRepeatedly(t *testing.T) { - t.Skip("This test still fails. Skip it until it is fixed.") - defer TearDownDataManagerTest(t) SetupDataManagerTest(t) @@ -526,10 +523,100 @@ func TestRunDatamanagerAsNonAdminUser(t *testing.T) { defer TearDownDataManagerTest(t) SetupDataManagerTest(t) - arv.ApiToken = ActiveUserToken + arv.ApiToken = arvadostest.ActiveToken err := singlerun(arv) if err == nil { t.Fatalf("Expected error during singlerun as non-admin user") } } + +// Create a collection with multiple streams and blocks +func createMultiStreamBlockCollection(t *testing.T, data string, numStreams, numBlocks int) (string, []string) { + defer switchToken(arvadostest.AdminToken)() + + manifest := "" + var locators []string + for s := 0; s < numStreams; s++ { + manifest += fmt.Sprintf("./stream%d ", s) + for b := 0; b < numBlocks; b++ { + locator, _, err := keepClient.PutB([]byte(fmt.Sprintf("%s in stream %d and block %d", data, s, b))) + if err != nil { + t.Fatalf("Error creating block %d in stream %d: %v", b, s, err) + } + locators = append(locators, strings.Split(locator, "+A")[0]) + manifest += locator + " " + } + manifest += "0:1:dummyfile.txt\n" + } + + collection := make(Dict) + err := arv.Create("collections", + arvadosclient.Dict{"collection": arvadosclient.Dict{"manifest_text": manifest}}, + &collection) + + if err != nil { + t.Fatalf("Error creating collection %v", err) + } + + return collection["uuid"].(string), locators +} + +/* + Create collection with multiple streams and blocks; backdate the blocks and but do not delete the collection. + Create another collection with multiple streams and blocks; backdate it's blocks and delete the collection. + After datamanager run: expect blocks from the first collection, but none from the second collection. +*/ +func TestPutAndGetCollectionsWithMultipleStreamsAndBlocks(t *testing.T) { + defer TearDownDataManagerTest(t) + SetupDataManagerTest(t) + + // create collection whose blocks will be backdated + collectionWithOldBlocks, oldBlocks := createMultiStreamBlockCollection(t, "old block", 100, 10) + if collectionWithOldBlocks == "" { + t.Fatalf("Failed to create collection with 1000 blocks") + } + if len(oldBlocks) != 1000 { + t.Fatalf("Not all blocks are created: expected %v, found %v", 1000, len(oldBlocks)) + } + + // create another collection, whose blocks will be backdated and the collection will be deleted + toBeDeletedCollection, toBeDeletedCollectionBlocks := createMultiStreamBlockCollection(t, "new block", 2, 5) + if toBeDeletedCollection == "" { + t.Fatalf("Failed to create collection with 10 blocks") + } + + // create a stray block that will be backdated + strayOldBlock := putBlock(t, "this stray block is old") + + // create another block that will not be backdated + strayNewBlock := putBlock(t, "this stray block is new") + + expected := []string{} + expected = append(expected, oldBlocks...) + expected = append(expected, toBeDeletedCollectionBlocks...) + expected = append(expected, strayOldBlock) + expected = append(expected, strayNewBlock) + verifyBlocks(t, nil, expected, 2) + + // Backdate old blocks; but the collection still references these blocks + backdateBlocks(t, oldBlocks) + + // Backdate first block from the newer blocks and delete the collection; the rest are still be reachable + backdateBlocks(t, toBeDeletedCollectionBlocks) + deleteCollection(t, toBeDeletedCollection) + + // also backdate the stray old block + backdateBlocks(t, []string{strayOldBlock}) + + // run datamanager + dataManagerSingleRun(t) + + expected = []string{strayNewBlock} + expected = append(expected, oldBlocks...) + + notExpected := []string{strayOldBlock} + notExpected = append(notExpected, toBeDeletedCollectionBlocks...) + + verifyBlocks(t, notExpected, expected, 2) +}