mongodb/mongo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

SERVER-126200 Remove chunkless references in the authoritative shard catalog (#54004)

Browse Source 
GitOrigin-RevId: 9f40ff9934f17e685c57eea190330327c380aadd
This commit is contained in:
Jordi Olivares Provencio 2026-05-22 00:40:41 +02:00 committed by MongoDB Bot
parent 231399c6e6
commit 11d71e49c2
17 changed files with 90 additions and 202 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
jstests/auth/internal_command_auth_validation.js
View File

@ -788,12 +788,14 @@ const internalCommandsMap = {
testname: "_shardsvrCommitRefineCollectionShardKey",
command: {
_shardsvrCommitRefineCollectionShardKey: "test.x",
primaryShardId: "",
},
},
_shardsvrCommitCollModCollectionMetadata: {
testname: "_shardsvrCommitCollModCollectionMetadata",
command: {
_shardsvrCommitCollModCollectionMetadata: "test.x",
primaryShardId: "",
},
},
_shardsvrCommitDropCollectionMetadata: {
@ -804,11 +806,17 @@ const internalCommandsMap = {
testname: "_shardsvrCommitCreateCollectionMetadata",
command: {
_shardsvrCommitCreateCollectionMetadata: "test.x",
primaryShardId: "",
},
},
_shardsvrCommitRenameCollectionMetadata: {
testname: "_shardsvrCommitRenameCollectionMetadata",
command: {_shardsvrCommitRenameCollectionMetadata: "", fromNss: "test.x", toNss: "test.y", primaryShardId: ""},
command: {
_shardsvrCommitRenameCollectionMetadata: "",
fromNss: "test.x",
toNss: "test.y",
primaryShardId: "",
},
},
_shardsvrSetAllowMigrations: {
testname: "_shardsvrSetAllowMigrations",

12
jstests/noPassthrough/global_catalog/authoritative_collection_metadata_vs_ddls.js
View File

@ -424,8 +424,7 @@ describe("Authoritative collection metadata vs DDLs", function () {
describe("convertToCapped", function () {
// For a convertToCapped on a tracked unsplittable collection: no node must retain chunks
// for the original UUID, the data shard must hold the new UUID's entry with real chunks,
// the DB primary must hold the new UUID's entry (chunkless if it is not the data shard),
// and any other shard must have no entry for the collection.
// the DB primary must hold the new UUID's entry, and any other shard must have no entry for the collection.
function assertShardCatalogAfterConvertToCapped(
ns,
{dataShardRs, primaryRs, newUuid, originalUuid, unsplittableKey},
@ -452,8 +451,7 @@ describe("Authoritative collection metadata vs DDLs", function () {
});
if (primaryRs !== dataShardRs) {
// Primary is not the data shard: it owns no real chunks but must carry a chunkless
// placeholder so disk recovery recognize the collection as tracked.
// Primary is not the data shard: it owns no real chunks but must carry the collection entry so disk recovery recognize the collection as tracked.
primaryRs.nodes.forEach((node) => {
const meta = getShardCatalogCollMetadata(node, ns);
assert.neq(null, meta, `${node.host}: chunkless primary is missing collection metadata`);
@ -464,11 +462,7 @@ describe("Authoritative collection metadata vs DDLs", function () {
);
const shardChunks = getShardCatalogChunks(node, newUuid);
assert.eq(
1,
shardChunks.length,
`${node.host}: chunkless primary must carry exactly one placeholder chunk`,
);
assert.eq(0, shardChunks.length, `${node.host}: chunkless primary must not have any chunks`);
});
}

16
src/mongo/db/global_catalog/ddl/collmod_coordinator.cpp
View File

@ -145,7 +145,6 @@ void commitToShardCatalog(OperationContext* opCtx,
const NamespaceString& nss,
const std::vector<ShardId>& participantsOwningChunks,
const ShardId& primaryShard,
bool isPrimaryOwningChunks,
GetSessionFn&& getSession,
const std::shared_ptr<executor::ScopedTaskExecutor>& executor,
const CancellationToken& token) {
@ -153,20 +152,14 @@ void commitToShardCatalog(OperationContext* opCtx,
return;
}
std::vector<ShardId> involvedShards = participantsOwningChunks;
if (isPrimaryOwningChunks) {
involvedShards.push_back(primaryShard);
auto shardIds = participantsOwningChunks;
if (std::find(shardIds.begin(), shardIds.end(), primaryShard) == shardIds.end()) {
shardIds.emplace_back(primaryShard);
}
const auto session = getSession();
sharding_ddl_util::commitCollModCollectionMetadataToShardCatalog(
opCtx, nss, involvedShards, session, executor, token);
// The DB primary shard must always know that a collection is tracked, even when it does not own
// any chunks.
if (!isPrimaryOwningChunks) {
shard_catalog_commit::commitChunklessCollectionLocally(opCtx, nss);
}
opCtx, nss, shardIds, session, executor, token);
}
} // namespace
@ -427,7 +420,6 @@ ExecutorFuture<void> CollModCoordinator::_runImpl(
_collInfo->nsForTargeting,
_shardingInfo->participantsOwningChunks,
_shardingInfo->primaryShard,
_shardingInfo->isPrimaryOwningChunks,
[&] { return getNewSession(opCtx); },
executor,
token);

22
src/mongo/db/global_catalog/ddl/convert_to_capped_coordinator.cpp
View File

@ -410,17 +410,19 @@ ExecutorFuture<void> ConvertToCappedCoordinator::_runImpl(
// Install the new incarnation of the collection in the shard catalog.
if (isAuthoritative) {
const auto session = getNewSession(opCtx);
sharding_ddl_util::commitCreateCollectionMetadataToShardCatalog(
opCtx, nss(), {*(_doc.getDataShard())}, session, executor, token);
// The DB primary shard must always know that the collection is tracked,
// even when it does not own any chunks. Persist a chunkless placeholder
// locally so that disk recovery can tell a chunkless-tracked collection
// apart from an untracked one.
const auto primaryShardId = ShardingState::get(opCtx)->shardId();
if (primaryShardId != *_doc.getDataShard()) {
shard_catalog_commit::commitChunklessCollectionLocally(opCtx, nss());
}
// even when it does not own any chunks.
std::set<ShardId> involvedShards;
involvedShards.emplace(ShardingState::get(opCtx)->shardId());
involvedShards.emplace(*(_doc.getDataShard()));
sharding_ddl_util::commitCreateCollectionMetadataToShardCatalog(
opCtx,
nss(),
{involvedShards.begin(), involvedShards.end()},
session,
executor,
token);
}
// Checkpoint the configTime to ensure that, in the case of a stepdown/crash,

12
src/mongo/db/global_catalog/ddl/create_collection_coordinator.cpp
View File

@ -2319,19 +2319,13 @@ void CreateCollectionCoordinator::_commitOnShardCatalog(
const auto& cm = uassertStatusOK(
Grid::get(opCtx)->catalogCache()->getCollectionPlacementInfoWithRefresh(opCtx, nss()));
cm.getAllShardIds(&involvedShards);
// The DB primary shard must always know that a collection is tracked, even when it does not
// own any chunks.
involvedShards.emplace(ShardingState::get(opCtx)->shardId());
const auto session = getNewSession(opCtx);
sharding_ddl_util::commitCreateCollectionMetadataToShardCatalog(
opCtx, nss(), {involvedShards.begin(), involvedShards.end()}, session, executor, token);
// The DB primary shard must always know that a collection is tracked, even when it does not
// own any chunks. We persist a placeholder chunk locally so that (1) disk recovery can
// distinguish a chunkless-tracked collection from an untracked one without special-case
// logic, and (2) CheckMetadataConsistency can verify the DB primary always has an entry.
const auto primaryShardId = ShardingState::get(opCtx)->shardId();
if (involvedShards.find(primaryShardId) == involvedShards.end()) {
shard_catalog_commit::commitChunklessCollectionLocally(opCtx, nss());
}
}
void CreateCollectionCoordinator::_setPostCommitMetadata(

14
src/mongo/db/global_catalog/ddl/refine_collection_shard_key_coordinator.cpp
View File

@ -342,20 +342,12 @@ ExecutorFuture<void> RefineCollectionShardKeyCoordinator::_runImpl(
if (_doc.getAuthoritativeMetadataAccessLevel() >=
AuthoritativeMetadataAccessLevelEnum::kWritesAllowed) {
const auto involvedShards = getShardsWithDataForCollection(opCtx, nss());
// The DB primary shard must always know that a collection is tracked, even when
// it does not own any chunks.
const auto involvedShards = getDataShardsAndDbPrimaryShard(opCtx, nss());
const auto session = getNewSession(opCtx);
sharding_ddl_util::commitRefineCollectionShardKeyToShardCatalog(
opCtx, nss(), involvedShards, session, executor, token);
// The DB primary shard must always know that a collection is tracked, even when
// it does not own any chunks. We persist a placeholder chunk locally so that
// disk recovery can distinguish a chunkless-tracked collection from an
// untracked one.
const auto primaryShardId = ShardingState::get(opCtx)->shardId();
if (std::find(involvedShards.begin(), involvedShards.end(), primaryShardId) ==
involvedShards.end()) {
shard_catalog_commit::commitChunklessCollectionLocally(opCtx, nss());
}
}
// Checkpoint the configTime to ensure that, in the case of a stepdown, the new

15
src/mongo/db/global_catalog/ddl/sharded_ddl_commands.idl
View File

@ -1104,6 +1104,11 @@ commands:
namespace: type
type: namespacestring
api_version: ""
fields:
primaryShardId:
type: shard_id
description: Primary shard identifier, used to determine whether the shard needs to durably persist collection metadata or not.
optional: false
reply_type: OkReply
_shardsvrCommitCollModCollectionMetadata:
@ -1114,6 +1119,11 @@ commands:
namespace: type
type: namespacestring
api_version: ""
fields:
primaryShardId:
type: shard_id
description: Primary shard identifier, used to determine whether the shard needs to durably persist collection metadata or not.
optional: false
reply_type: OkReply
_shardsvrCommitDropCollectionMetadata:
@ -1138,6 +1148,11 @@ commands:
namespace: type
type: namespacestring
api_version: ""
fields:
primaryShardId:
type: shard_id
description: Primary shard identifier, used to determine whether the shard needs to durably persist collection metadata or not.
optional: false
reply_type: OkReply
_shardsvrCommitRenameCollectionMetadata:

6
src/mongo/db/global_catalog/ddl/sharding_ddl_util.cpp
View File

@ -893,7 +893,7 @@ void commitRefineCollectionShardKeyToShardCatalog(
const OperationSessionInfo& osi,
const std::shared_ptr<executor::ScopedTaskExecutor>& executor,
const CancellationToken& token) {
ShardsvrCommitRefineCollectionShardKey request(nss);
ShardsvrCommitRefineCollectionShardKey request(nss, ShardingState::get(opCtx)->shardId());
request.setDbName(DatabaseName::kAdmin);
generic_argument_util::setMajorityWriteConcern(request);
@ -913,7 +913,7 @@ void commitCollModCollectionMetadataToShardCatalog(
const OperationSessionInfo& osi,
const std::shared_ptr<executor::ScopedTaskExecutor>& executor,
const CancellationToken& token) {
ShardsvrCommitCollModCollectionMetadata request(nss);
ShardsvrCommitCollModCollectionMetadata request(nss, ShardingState::get(opCtx)->shardId());
request.setDbName(DatabaseName::kAdmin);
generic_argument_util::setMajorityWriteConcern(request);
@ -954,7 +954,7 @@ void commitCreateCollectionMetadataToShardCatalog(
const OperationSessionInfo& osi,
const std::shared_ptr<executor::ScopedTaskExecutor>& executor,
const CancellationToken& token) {
ShardsvrCommitCreateCollectionMetadata request(nss);
ShardsvrCommitCreateCollectionMetadata request(nss, ShardingState::get(opCtx)->shardId());
request.setDbName(DatabaseName::kAdmin);
generic_argument_util::setMajorityWriteConcern(request);

61
src/mongo/db/shard_role/shard_catalog/commit_collection_metadata_locally.cpp
View File

@ -293,38 +293,6 @@ void clearShardCatalogCacheForDroppedCollection(OperationContext* opCtx,
} // namespace
void commitRefineShardKeyLocally(OperationContext* opCtx, const NamespaceString& nss) {
auto coll = fetchCollection(opCtx, nss);
auto ownedChunks = fetchOwnedChunks(opCtx, nss, coll);
// Write to `config.shard.catalog.(collections|chunks)` to insert collection metadata.
writeCollectionMetadataLocally(opCtx, nss, coll.asShardCatalogType(), ownedChunks);
// Delete stale chunks from config.shard.catalog.chunks whose shard key bounds do not match the
// refined key pattern. This can occur when the shard catalog has an out-of-date view of the
// owned chunk ranges (e.g., due to splits or merges).
// TODO (SERVER-121709): Evaluate if this holds once merge/split are authoritative.
const int numKeyFields = coll.getKeyPattern().toBSON().nFields();
{
DBDirectClient dbClient(opCtx);
executeLocalDelete(
dbClient,
NamespaceString::kConfigShardCatalogChunksNamespace,
BSON(ChunkType::collectionUUID()
<< coll.getUuid() << "$expr"
<< BSON("$ne" << BSON_ARRAY(BSON("$size" << BSON("$objectToArray" << "$min"))
<< numKeyFields))),
true /* multi */);
}
// Write an oplog 'c' entry to invalidate collection metadata on secondaries.
invalidateCollectionMetadataOnSecondaries(
opCtx, nss, coll.getUuid(), false /* forDroppedCollection */);
// Update this node CSR with collection metadata and chunks.
updateShardCatalogCache(opCtx, nss, coll, ownedChunks);
}
void commitDropCollectionLocally(OperationContext* opCtx,
const NamespaceString& nss,
const UUID& uuid) {
@ -480,34 +448,5 @@ void commitCollectionMetadataLocally(OperationContext* opCtx,
opCtx, nss, coll.getUuid(), false /* forDroppedCollection */);
}
void commitChunklessCollectionLocally(OperationContext* opCtx, const NamespaceString& nss) {
auto coll = fetchCollection(opCtx, nss);
// Drop all existing chunk entries for this collection to start from a clean slate. This removes
// both real chunks and any prior chunkless placeholder.
deleteCollectionChunksMetadataLocally(opCtx, coll.getUuid());
// This shard does not own any chunks, but we still need the CSS to know the collection is
// tracked. Persist a single placeholder chunk so that disk recovery can distinguish a
// chunkless-tracked collection from an untracked one without special-case logic.
auto range = ChunkRange(coll.getKeyPattern().globalMin(), coll.getKeyPattern().globalMax());
ChunkType placeholder(coll.getUuid(),
std::move(range),
ChunkVersion({coll.getEpoch(), coll.getTimestamp()}, {1, 0}),
kChunklessPlaceholderShardId);
placeholder.setName(OID::gen());
std::vector<ChunkType> placeholderChunks{std::move(placeholder)};
// Write the collection document and the placeholder chunk to the shard catalog.
writeCollectionMetadataLocally(opCtx, nss, coll.asShardCatalogType(), placeholderChunks);
// Write an oplog 'c' entry to invalidate collection metadata on secondaries.
invalidateCollectionMetadataOnSecondaries(
opCtx, nss, coll.getUuid(), false /* forDroppedCollection */);
// Update this node CSR with chunkless tracked metadata.
updateShardCatalogCache(opCtx, nss, coll, placeholderChunks);
}
} // namespace shard_catalog_commit
} // namespace mongo

23
src/mongo/db/shard_role/shard_catalog/commit_collection_metadata_locally.h
View File

@ -38,20 +38,6 @@ MONGO_MOD_PARENT_PRIVATE;
namespace mongo {
namespace shard_catalog_commit {
/**
* Shard id for the placeholder chunk when a collection is tracked on a shard but owns no real
* chunks (see commitCreateCollectionChunklessLocally).
*/
inline const ShardId kChunklessPlaceholderShardId{"__chunkless_placeholder__"};
/**
* Fetches the latest collection metadata and owned chunks from the global catalog, persists them
* to the shard catalog (config.shard.catalog.collections and config.shard.catalog.chunks), removes
* any stale chunks whose shard key bounds don't match the refined key pattern, writes an oplog
* entry to invalidate collection metadata on secondaries, and updates the in-memory
* CollectionShardingRuntime (CSR) with the new routing information.
*/
void commitRefineShardKeyLocally(OperationContext* opCtx, const NamespaceString& nss);
/**
* Deletes the collection and chunk metadata from the shard catalog
@ -99,14 +85,5 @@ void commitCollectionMetadataLocally(OperationContext* opCtx,
const NamespaceString& nss,
bool isDbPrimaryShard = false);
/**
* Fetches the collection metadata from the global catalog, removes any existing chunk entries for
* this collection UUID, persists a chunkless placeholder chunk and the updated collection document
* to the shard catalog, writes an oplog entry to invalidate secondaries, and refreshes the
* in-memory CollectionShardingRuntime. Used for shards that participate in a tracked collection
* but own no chunks.
*/
void commitChunklessCollectionLocally(OperationContext* opCtx, const NamespaceString& nss);
} // namespace shard_catalog_commit
} // namespace mongo

48
src/mongo/db/shard_role/shard_catalog/commit_collection_metadata_locally_test.cpp
View File

@ -184,7 +184,7 @@ TEST_F(CommitCollectionMetadataLocallyTest, RefineShardKeyPersistsCollectionAndC
auto [collType, chunks] = makeCollectionMetadata(3);
mockCatalogClient()->setCollectionMetadata(collType, chunks);
shard_catalog_commit::commitRefineShardKeyLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogCollectionsNamespace), 1);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 3);
@ -201,7 +201,7 @@ TEST_F(CommitCollectionMetadataLocallyTest, RefineShardKeyUpdatesCSR) {
auto [collType, chunks] = makeCollectionMetadata(2);
mockCatalogClient()->setCollectionMetadata(collType, chunks);
shard_catalog_commit::commitRefineShardKeyLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
auto scopedCsr = CollectionShardingRuntime::acquireShared(operationContext(), kTestNss);
auto metadata = scopedCsr->getCurrentMetadataIfKnown();
@ -214,8 +214,8 @@ TEST_F(CommitCollectionMetadataLocallyTest, RefineShardKeyIsIdempotent) {
auto [collType, chunks] = makeCollectionMetadata(2);
mockCatalogClient()->setCollectionMetadata(collType, chunks);
shard_catalog_commit::commitRefineShardKeyLocally(operationContext(), kTestNss);
shard_catalog_commit::commitRefineShardKeyLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogCollectionsNamespace), 1);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 2);
@ -301,7 +301,7 @@ TEST_F(CommitCollectionMetadataLocallyTest, ChunklessCollectionPersistsTokenToDi
auto [collType, chunks] = makeCollectionMetadata(0);
mockCatalogClient()->setCollectionMetadata(collType, {});
shard_catalog_commit::commitChunklessCollectionLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogCollectionsNamespace), 1);
@ -309,18 +309,15 @@ TEST_F(CommitCollectionMetadataLocallyTest, ChunklessCollectionPersistsTokenToDi
ASSERT_EQ(collDocs.size(), 1u);
ASSERT_EQ(UUID::fromCDR(collDocs[0].getField("uuid").uuid()), collType.getUuid());
// The placeholder chunk is persisted so the token survives restarts.
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 1);
auto chunkDocs = findLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace);
ASSERT_EQ(UUID::fromCDR(chunkDocs[0].getField(ChunkType::collectionUUID.name()).uuid()),
collType.getUuid());
// No chunk should be stored in the local chunks for a chunkless collection on the dbPrimary.
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 0);
}
TEST_F(CommitCollectionMetadataLocallyTest, ChunklessCollectionUpdatesCSR) {
auto [collType, chunks] = makeCollectionMetadata(0);
mockCatalogClient()->setCollectionMetadata(collType, {});
shard_catalog_commit::commitChunklessCollectionLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
auto scopedCsr = CollectionShardingRuntime::acquireShared(operationContext(), kTestNss);
auto metadata = scopedCsr->getCurrentMetadataIfKnown();
@ -335,23 +332,21 @@ TEST_F(CommitCollectionMetadataLocallyTest, ChunklessCollectionIsIdempotent) {
auto [collType, _] = makeCollectionMetadata(0);
mockCatalogClient()->setCollectionMetadata(collType, {});
shard_catalog_commit::commitChunklessCollectionLocally(operationContext(), kTestNss);
shard_catalog_commit::commitChunklessCollectionLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
// Repeated calls must not accumulate placeholder rows; each call generates a fresh OID, so
// this only holds if the helper deletes the prior placeholder before inserting.
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogCollectionsNamespace), 1);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 1);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 0);
}
TEST_F(CommitCollectionMetadataLocallyTest, RefineShardKeyChunklessPersistsCollectionWithNewEpoch) {
// Seed a chunkless tracked collection at (epoch1, ts1).
auto [collType1, _] = makeCollectionMetadata(0);
mockCatalogClient()->setCollectionMetadata(collType1, {});
shard_catalog_commit::commitChunklessCollectionLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogCollectionsNamespace), 1);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 1);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 0);
// Simulate a refine to (epoch2, ts2) on the same UUID with an extended key pattern.
const OID epoch2 = OID::gen();
@ -361,18 +356,13 @@ TEST_F(CommitCollectionMetadataLocallyTest, RefineShardKeyChunklessPersistsColle
kTestNss, epoch2, ts2, Date_t::now(), collType1.getUuid(), newKeyPattern};
mockCatalogClient()->setCollectionMetadata(collType2, {});
shard_catalog_commit::commitChunklessCollectionLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
// The collection doc reflects the new triple and exactly one placeholder chunk persists.
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogCollectionsNamespace), 1);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 1);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 0);
auto collDocs = findLocalDocs(NamespaceString::kConfigShardCatalogCollectionsNamespace);
ASSERT_BSONOBJ_EQ(collDocs[0].getObjectField("key"), newKeyPattern);
auto chunkDocs = findLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace);
ASSERT_EQ(UUID::fromCDR(chunkDocs[0].getField(ChunkType::collectionUUID.name()).uuid()),
collType1.getUuid());
}
TEST_F(CommitCollectionMetadataLocallyTest, RefineShardKeyRemovesStaleChunks) {
@ -400,7 +390,7 @@ TEST_F(CommitCollectionMetadataLocallyTest, RefineShardKeyRemovesStaleChunks) {
}
mockCatalogClient()->setCollectionMetadata(collType, chunks);
shard_catalog_commit::commitRefineShardKeyLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss, true);
// The stale chunks should have been removed and replaced with the correct ones.
auto chunkDocs = findLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace);
@ -415,7 +405,7 @@ TEST_F(CommitCollectionMetadataLocallyTest, DropCollectionDeletesMetadata) {
mockCatalogClient()->setCollectionMetadata(collType, chunks);
// First commit the metadata so there's something to drop.
shard_catalog_commit::commitRefineShardKeyLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogCollectionsNamespace), 1);
ASSERT_EQ(countLocalDocs(NamespaceString::kConfigShardCatalogChunksNamespace), 3);
@ -430,7 +420,7 @@ TEST_F(CommitCollectionMetadataLocallyTest, DropCollectionClearsCSR) {
auto [collType, chunks] = makeCollectionMetadata(2);
mockCatalogClient()->setCollectionMetadata(collType, chunks);
shard_catalog_commit::commitRefineShardKeyLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss);
{
auto scopedCsr = CollectionShardingRuntime::acquireShared(operationContext(), kTestNss);
@ -456,7 +446,7 @@ TEST_F(CommitCollectionMetadataLocallyTest, DropCollectionIsNoOpOnEmptyCatalog)
TEST_F(CommitCollectionMetadataLocallyTest, DropCollectionOnlyDeletesTargetCollection) {
auto [collType1, chunks1] = makeCollectionMetadata(2);
mockCatalogClient()->setCollectionMetadata(collType1, chunks1);
shard_catalog_commit::commitRefineShardKeyLocally(operationContext(), kTestNss);
shard_catalog_commit::commitCollectionMetadataLocally(operationContext(), kTestNss);
// Insert a second collection's data directly.
auto otherNss = NamespaceString::createNamespaceString_forTest("TestDB", "OtherColl");

12
src/mongo/db/shard_role/shard_catalog/shard_catalog_history_cleanup.cpp
View File

@ -58,17 +58,13 @@ StorageEngine::TimestampMonitor::TimestampListener kShardCatalogHistoryCleanupTi
auto shardId = shardingState->shardId();
// TODO(SERVER-126200): Remove reference to ChunklessPlaceholder.
PersistentTaskStore<ChunkType> chunkStore{
NamespaceString::kConfigShardCatalogChunksNamespace};
try {
chunkStore.remove(
opCtx,
BSON(ChunkType::shard()
<< BSON("$nin" << BSON_ARRAY(
shardId.toString()
<< shard_catalog_commit::kChunklessPlaceholderShardId.toString()))
<< ChunkType::onCurrentShardSince() << BSON("$lt" << oldest)));
chunkStore.remove(opCtx,
BSON(ChunkType::shard()
<< BSON("$ne" << shardId.toString())
<< ChunkType::onCurrentShardSince() << BSON("$lt" << oldest)));
} catch (const ExceptionFor<ErrorCodes::FailedToSatisfyReadPreference>&) {
// Primary can be killed in the middle of the removal.
return;

16
src/mongo/db/shard_role/shard_catalog/shard_catalog_history_cleanup_test.cpp
View File

@ -77,7 +77,6 @@ public:
TEST_F(ShardCatalogHistoryCleanupTest, ShardCatalogHistoryCleanupCalledOnTimestampMonitorAdvance) {
const ShardId kShardName{"testShard"};
const ShardId kAnotherShardName{"anotherTestShard"};
const ShardId kChunklessPlaceholderShardId{"__chunkless_placeholder__"};
// Imagining the test is running on a primary in sharded cluster
serverGlobalParams.clusterRole = ClusterRole::ShardServer;
@ -133,13 +132,6 @@ TEST_F(ShardCatalogHistoryCleanupTest, ShardCatalogHistoryCleanupCalledOnTimesta
notOutdatedPreviouslyOwnedChunkType.setHistory(
{ChunkHistory(newTimestamp, kAnotherShardName), ChunkHistory(oldestTimestamp, kShardName)});
notOutdatedPreviouslyOwnedChunkType.setOnCurrentShardSince(newTimestamp);
// Note: in reality placeholder would always store MINKEY, MAXKEY range, but in this unit test
// we can disregard it
auto placehoderOwnedChunkType =
ChunkType(collectionUUID,
ChunkRange(BSON("_id" << 300), BSON("_id" << MAXKEY)),
ChunkVersion(CollectionGeneration{epoch, Timestamp(0, 0)}, placement),
kChunklessPlaceholderShardId);
DBDirectClient client(opCtx.get());
client.insert(NamespaceString::kConfigShardCatalogChunksNamespace,
@ -148,8 +140,6 @@ TEST_F(ShardCatalogHistoryCleanupTest, ShardCatalogHistoryCleanupCalledOnTimesta
outdatedPreviouslyOwnedChunkType.toConfigBSON());
client.insert(NamespaceString::kConfigShardCatalogChunksNamespace,
notOutdatedPreviouslyOwnedChunkType.toConfigBSON());
client.insert(NamespaceString::kConfigShardCatalogChunksNamespace,
placehoderOwnedChunkType.toConfigBSON());
waitForTimestampMonitorPass();
waitForTimestampMonitorPass();
@ -164,11 +154,11 @@ TEST_F(ShardCatalogHistoryCleanupTest, ShardCatalogHistoryCleanupCalledOnTimesta
chunks.push_back(chunkRes.getValue());
}
ASSERT_EQ(chunks.size(), 3u);
ASSERT_EQ(chunks.size(), 2u);
// Note: Parsing from ChunkType overrode the chunk version, so we have to set it manually
std::vector<ChunkType> expectedChunks = {
currentlyOwnedChunkType, notOutdatedPreviouslyOwnedChunkType, placehoderOwnedChunkType};
std::vector<ChunkType> expectedChunks = {currentlyOwnedChunkType,
notOutdatedPreviouslyOwnedChunkType};
for (auto& chunk : expectedChunks) {
chunk.setVersion(ChunkVersion(CollectionGeneration(epoch, newTimestamp), placement));
}

10
src/mongo/db/shard_role/shard_catalog/shard_filtering_metadata_refresh_test.cpp
View File

@ -999,21 +999,11 @@ TEST_F(AuthoritativeRefreshFixture, TrackedCollectionWithNoChunksOnDiskRecovered
const Timestamp timestamp(Date_t::now());
CollectionType collType{kTestNss, epoch, timestamp, Date_t::now(), uuid, kShardKeyPattern};
auto keyPattern = KeyPattern(kShardKeyPattern);
auto range = ChunkRange(keyPattern.globalMin(), keyPattern.globalMax());
ChunkType placeholder(uuid,
std::move(range),
ChunkVersion({epoch, timestamp}, {1, 0}),
shard_catalog_commit::kChunklessPlaceholderShardId);
placeholder.setName(OID::gen());
createTestCollection(opCtx, NamespaceString::kConfigShardCatalogCollectionsNamespace);
createTestCollection(opCtx, NamespaceString::kConfigShardCatalogChunksNamespace);
{
DBDirectClient client(opCtx);
client.insert(NamespaceString::kConfigShardCatalogCollectionsNamespace, collType.toBSON());
client.insert(NamespaceString::kConfigShardCatalogChunksNamespace,
placeholder.toConfigBSON());
}
{

5
src/mongo/db/shard_role/shard_catalog/shardsvr_commit_collmod_collection_metadata_command.cpp
View File

@ -95,7 +95,10 @@ public:
Grid::get(opCtx->getServiceContext())->getExecutorPool()->getFixedExecutor());
newOpCtx->setAlwaysInterruptAtStepDownOrUp_UNSAFE();
shard_catalog_commit::commitCollectionMetadataLocally(newOpCtx.get(), ns());
bool isPrimaryDbShard =
ShardingState::get(newOpCtx.get())->shardId() == request().getPrimaryShardId();
shard_catalog_commit::commitCollectionMetadataLocally(
newOpCtx.get(), ns(), isPrimaryDbShard);
}
LOGV2_INFO(12591501,

5
src/mongo/db/shard_role/shard_catalog/shardsvr_commit_create_collection_metadata_command.cpp
View File

@ -95,7 +95,10 @@ public:
Grid::get(opCtx->getServiceContext())->getExecutorPool()->getFixedExecutor());
newOpCtx->setAlwaysInterruptAtStepDownOrUp_UNSAFE();
shard_catalog_commit::commitCollectionMetadataLocally(newOpCtx.get(), ns());
bool isPrimaryDbShard =
ShardingState::get(newOpCtx.get())->shardId() == request().getPrimaryShardId();
shard_catalog_commit::commitCollectionMetadataLocally(
newOpCtx.get(), ns(), isPrimaryDbShard);
}
LOGV2_INFO(

5
src/mongo/db/shard_role/shard_catalog/shardsvr_commit_refine_collection_shard_key_command.cpp
View File

@ -94,7 +94,10 @@ public:
Grid::get(opCtx->getServiceContext())->getExecutorPool()->getFixedExecutor());
newOpCtx->setAlwaysInterruptAtStepDownOrUp_UNSAFE();
shard_catalog_commit::commitRefineShardKeyLocally(newOpCtx.get(), nss);
bool isPrimaryDbShard =
ShardingState::get(newOpCtx.get())->shardId() == request().getPrimaryShardId();
shard_catalog_commit::commitCollectionMetadataLocally(
newOpCtx.get(), nss, isPrimaryDbShard);
}
LOGV2_INFO(10281504,