Cherry-pick from master
pr: #45455
Related to #45445
Previously, FillFieldData for JSON fields would assert and fail when a
default_value was provided, blocking index creation for JSON fields with
default values (including dynamic fields like $meta).
This change enables JSON default value support by:
- Removing the assertion that blocked default values
- Parsing bytes_data into Json objects when default_value is present
- Properly filling data_ array and setting valid_data_ bitset to true
- Maintaining null behavior when no default_value is provided
Impact:
- Fixes index creation failure for JSON fields with default values
- Resolves upgrade issues from 2.5 to 2.6.5 where dynamic fields with
default values couldn't be indexed
- Index builds that were stuck in InProgress state can now complete
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
Cherry-pick from master
pr: #45444
Related to #45338
When using bulk vector search in hybrid search with rerank functions,
the output field values for different queries were all equal to the
values returned by the first query, instead of the correct values
belonging to each document ID. The document IDs were correct, but the
entity field values were wrong.
In rerank functions (RRF, weighted, decay, model), when processing
multiple queries in a batch, the `idLocations` stored only the relative
offset within each result set (`idx`), not accounting for the absolute
position within the entire batch. This caused `FillFieldData` to
retrieve field data from the wrong positions, always using offsets
relative to the first query.
This fix ensures that when processing bulk searches with rerank
functions, each result correctly retrieves its corresponding field data
based on the absolute offset within the entire batch, resolving the
issue where all queries returned the first query's field values.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
Sliced arrow arrays "incorrectly" returned the original array's size via
SizeInBytes(), causing inaccurate memory estimates during compaction.
This resulted in segments closing prematurely in mergeSplit mode -
expected 500MB compactions produced 4x100+MB segments instead.
Fixed by calculating actual byte size of sliced arrays, ensuring proper
segment sizing and more accurate memory usage tracking.
See also: #45293
pr: #45294
Signed-off-by: yangxuan <xuan.yang@zilliz.com>
When there're no growing segments in the collection, L0 Compaction will
try to choose all L0 segments that hits all L1/L2 segments.
However, if there's Sealed Segment still under flushing in DataNode at
the same time L0 Compaction selects satisfied L1/L2 segments, L0
Compaction will ignore this Segment because it's not in "FlushState",
which is wrong, causing missing deletes on the Sealed Segment.
This quick solution here is to fail this L0 compaction task once
selected a Sealed segment.
See also: #45339
pr: #45340
---------
Signed-off-by: yangxuan <xuan.yang@zilliz.com>
Cherry-pick from master
pr: #45334
Related to #45333
Fix segment loading failure when adding fields with text match enabled.
The issue occurred because text indexes were being loaded before
FinishLoad() was called, meaning raw data was not properly available
when text index creation attempted to access it, resulting in "failed to
create text index, neither raw data nor index are found" errors.
Solution is to move the FinishLoad() call to execute after raw data
loading but before text index loading. This ensures that:
1. Raw data is properly loaded and available in memory
2. Text indexes can access the raw data they need during creation
3. The segment is in the correct state before any index operations
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
Cherry-pick from master
pr: #45330
Related to #45329
Fix compaction failure when handling newly added dynamic fields with
storage v1 binlogs. The issue occurred because the
`GenerateEmptyArrayFromSchema` function did not support JSON data type
default values, causing "Unexpected default value" errors during
compaction.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #45080, #45274, #45285
pr: #45290
- LoadCollection doesn't ignore the ignorable request, for false field
array.
- CreatIndex doesn't ignore the ignorable request, for wrong index.
- index meta is not thread safe.
- lost parameter check of DDL.
- DDL Ack scheduler may get stuck and DDL is block until next incoming
DDL.
Signed-off-by: chyezh <chyezh@outlook.com>
issue: #43897, #44123
pr: #45266
also pick pr: #45237, #45264,#45244,#45275
fix: kafka should auto reset the offset from earliest to read (#45237)
issue: #44172, #45210, #44851,#45244
kafka will auto reset the offset to "latest" if the offset is
Out-of-range. the recovery of milvus wal cannot read any message from
that. So once the offset is out-of-range, kafka should read from eariest
to read the latest uncleared data.
https://kafka.apache.org/documentation/#consumerconfigs_auto.offset.reset
enhance: support alter collection/database with WAL-based DDL framework
(#45266)
issue: #43897
- Alter collection/database is implemented by WAL-based DDL framework
now.
- Support AlterCollection/AlterDatabase in wal now.
- Alter operation can be synced by new CDC now.
- Refactor some UT for alter DDL.
fix: milvus role cannot stop at initializing state (#45244)
issue: #45243
fix: support upgrading from 2.6.x -> 2.6.5 (#45264)
issue: #43897
---------
Signed-off-by: chyezh <chyezh@outlook.com>
Related to #45282
Initialize `tsFrom` and `tsTo` fields in the composite binlog record
writer constructor to prevent timestamp range information loss in stats
tasks.
The composite binlog writer now properly initializes the timestamp range
fields, ensuring that:
1. The first timestamp update will correctly set the minimum (`tsFrom`)
2. The first timestamp update will correctly set the maximum (`tsTo`)
3. All subsequent data writes will maintain accurate timestamp range
tracking
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
Cherry pick from master
pr: #45263
Related to #43028
Initialize the schema version field when creating a new collection
instance in QueryNode. The schema version is extracted from loadMetaInfo
and assigned to the collection, ensuring proper schema version tracking
and consistency across the distributed system.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #43897, #44123
pr: #45224
also pick pr: #45216,#45154,#45033,#45145,#45092,#45058,#45029
enhance: Close channel replicator more gracefully (#45029)
issue: https://github.com/milvus-io/milvus/issues/44123
enhance: Show create time for import job (#45058)
issue: https://github.com/milvus-io/milvus/issues/45056
fix: wal state may be unconsistent after recovering from crash (#45092)
issue: #45088, #45086
- Message on control channel should trigger the checkpoint update.
- LastConfrimedMessageID should be recovered from the minimum of
checkpoint or the LastConfirmedMessageID of uncommitted txn.
- Add more log info for wal debugging.
fix: make ack of broadcaster cannot canceled by client (#45145)
issue: #45141
- make ack of broadcaster cannot canceled by rpc.
- make clone for assignment snapshot of wal balancer.
- add server id for GetReplicateCheckpoint to avoid failure.
enhance: support collection and index with WAL-based DDL framework
(#45033)
issue: #43897
- Part of collection/index related DDL is implemented by WAL-based DDL
framework now.
- Support following message type in wal, CreateCollection,
DropCollection, CreatePartition, DropPartition, CreateIndex, AlterIndex,
DropIndex.
- Part of collection/index related DDL can be synced by new CDC now.
- Refactor some UT for collection/index DDL.
- Add Tombstone scheduler to manage the tombstone GC for collection or
partition meta.
- Move the vchannel allocation into streaming pchannel manager.
enhance: support load/release collection/partition with WAL-based DDL
framework (#45154)
issue: #43897
- Load/Release collection/partition is implemented by WAL-based DDL
framework now.
- Support AlterLoadConfig/DropLoadConfig in wal now.
- Load/Release operation can be synced by new CDC now.
- Refactor some UT for load/release DDL.
enhance: Don't start cdc by default (#45216)
issue: https://github.com/milvus-io/milvus/issues/44123
fix: unrecoverable when replicate from old (#45224)
issue: #44962
---------
Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
Signed-off-by: chyezh <chyezh@outlook.com>
Co-authored-by: yihao.dai <yihao.dai@zilliz.com>
issue: #43897, #44123
pr: #44898
related pr: #44607#44642#44792#44809#44564#44560#44735#44822#44865#44850#44942#44874#44963#44886#44898
enhance: remove redundant channel manager from datacoord (#44532)
issue: #41611
- After enabling streaming arch, channel manager of data coord is a
redundant component.
fix: Fix CDC OOM due to high buffer size (#44607)
Fix CDC OOM by:
1. free msg buffer manually.
2. limit max msg buffer size.
3. reduce scanner msg hander buffer size.
issue: https://github.com/milvus-io/milvus/issues/44123
fix: remove wrong start timetick to avoid filtering DML whose timetick
is less than it. (#44691)
issue: #41611
- introduced by #44532
enhance: support remove cluster from replicate topology (#44642)
issue: #44558, #44123
- Update config(A->C) to A and C, config(B) to B on replicate topology
(A->B,A->C) can remove the B from replicate topology
- Fix some metric error of CDC
fix: check if qn is sqn with label and streamingnode list (#44792)
issue: #44014
- On standalone, the query node inside need to load segment and watch
channel, so the querynode is not a embeded querynode in streamingnode
without `LabelStreamingNodeEmbeddedQueryNode`. The channel dist manager
can not confirm a standalone node is a embededStreamingNode.
Bug is introduced by #44099
enhance: Make GetReplicateInfo API work at the pchannel level (#44809)
issue: https://github.com/milvus-io/milvus/issues/44123
enhance: Speed up CDC scheduling (#44564)
Make CDC watch etcd replicate pchannel meta instead of listing them
periodically.
issue: https://github.com/milvus-io/milvus/issues/44123
enhance: refactor update replicate config operation using
wal-broadcast-based DDL/DCL framework (#44560)
issue: #43897
- UpdateReplicateConfig operation will broadcast AlterReplicateConfig
message into all pchannels with cluster-exclusive-lock.
- Begin txn message will use commit message timetick now (to avoid
timetick rollback when CDC with txn message).
- If current cluster is secondary, the UpdateReplicateConfig will wait
until the replicate configuration is consistent with the config
replicated from primary.
enhance: support rbac with WAL-based DDL framework (#44735)
issue: #43897
- RBAC(Roles/Users/Privileges/Privilege Groups) is implemented by
WAL-based DDL framework now.
- Support following message type in wal `AlterUser`, `DropUser`,
`AlterRole`, `DropRole`, `AlterUserRole`, `DropUserRole`,
`AlterPrivilege`, `DropPrivilege`, `AlterPrivilegeGroup`,
`DropPrivilegeGroup`, `RestoreRBAC`.
- RBAC can be synced by new CDC now.
- Refactor some UT for RBAC.
enhance: support database with WAL-based DDL framework (#44822)
issue: #43897
- Database related DDL is implemented by WAL-based DDL framework now.
- Support following message type in wal CreateDatabase, AlterDatabase,
DropDatabase.
- Database DDL can be synced by new CDC now.
- Refactor some UT for Database DDL.
enhance: support alias with WAL-based DDL framework (#44865)
issue: #43897
- Alias related DDL is implemented by WAL-based DDL framework now.
- Support following message type in wal AlterAlias, DropAlias.
- Alias DDL can be synced by new CDC now.
- Refactor some UT for Alias DDL.
enhance: Disable import for replicating cluster (#44850)
1. Import in replicating cluster is not supported yet, so disable it for
now.
2. Remove GetReplicateConfiguration wal API
issue: https://github.com/milvus-io/milvus/issues/44123
fix: use short debug string to avoid newline in debug logs (#44925)
issue: #44924
fix: rerank before requery if reranker didn't use field data (#44942)
issue: #44918
enhance: support resource group with WAL-based DDL framework (#44874)
issue: #43897
- Resource group related DDL is implemented by WAL-based DDL framework
now.
- Support following message type in wal AlterResourceGroup,
DropResourceGroup.
- Resource group DDL can be synced by new CDC now.
- Refactor some UT for resource group DDL.
fix: Fix Fix replication txn data loss during chaos (#44963)
Only confirm CommitMsg for txn messages to prevent data loss.
issue: https://github.com/milvus-io/milvus/issues/44962,
https://github.com/milvus-io/milvus/issues/44123
fix: wrong execution order of DDL/DCL on secondary (#44886)
issue: #44697, #44696
- The DDL executing order of secondary keep same with order of control
channel timetick now.
- filtering the control channel operation on shard manager of
streamingnode to avoid wrong vchannel of create segment.
- fix that the immutable txn message lost replicate header.
fix: Fix primary-secondary replication switch blocking (#44898)
1. Fix primary-secondary replication switchover blocking by delete
replicate pchannel meta using modRevision.
2. Stop channel replicator(scanner) when cluster role changes to prevent
continued message consumption and replication.
3. Close Milvus client to prevent goroutine leak.
4. Create Milvus client once for a channel replicator.
5. Simplify CDC controller and resources.
issue: https://github.com/milvus-io/milvus/issues/44123
---------
Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
Signed-off-by: chyezh <chyezh@outlook.com>
Co-authored-by: yihao.dai <yihao.dai@zilliz.com>
issue: #45077
pr: #45078
We need to promise the state of wal consistent with the memory state of
streamingnode. So we don't allow the append operation can be cancelled
by the append caller to avoid leave a inconsistent state of alive wal.
The wal append operation can only be cancelled when the wal is shutting
down.
Signed-off-by: chyezh <chyezh@outlook.com>
Cherry-pick from master
pr: #45158
Related to #45157
Fix a bug where DataNode panics when building json stats index throws an
exception before the writer is initialized. The destructor would call
Close() on an uninitialized packed_writer_ pointer, causing a null
pointer dereference.
Changes:
- Add null check for packed_writer_ before calling Flush() and Close()
- Prevents null pointer dereference in edge cases
- Ignore close status as this is a cleanup operation
This ensures safe cleanup even when initialization fails due to
exceptions.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
Cherry-pick from master
pr: #45147
Related to #44509
Fix a bug where QueryNodeNumEntities metrics were not updated for
collections with zero segments, causing stale metrics when all segments
are flushed or compacted.
The previous implementation used separate loops: one to update size
metrics for all collections, and another to update num entities metrics
only for collections present in the grouped segments map. Collections
with no segments were skipped in the second loop, leaving their
NumEntities metrics stale.
Changes:
- Consolidate size and num entities metric updates into single loop
- Iterate over all collections instead of grouped segments
- Get collection metadata from manager instead of segment instances
- Correctly set NumEntities to 0 for collections with no segments
- Apply the same fix to both growing and sealed segment processing
- Add nil check for collection metadata before processing
This ensures all collection metrics are updated consistently, even when
segment count drops to zero.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #44909
pr: #44917
When requery optimization is enabled, search results contain IDs but
empty FieldsData. During reduce/rerank operations, if the first shard
has empty FieldsData while others have data, PrepareResultFieldData
initializes an empty array, causing AppendFieldData to panic when
accessing array indices.
Changes:
- Find first non-empty FieldsData as template in 3 functions:
reduceAdvanceGroupBy, reduceSearchResultDataWithGroupBy,
reduceSearchResultDataNoGroupBy
- Add length check before 2 AppendFieldData calls in reduce functions to
prevent panic
- Improve newRerankOutputs to find first non-empty fieldData using
len(FieldsData) check instead of GetSizeOfIDs
- Add length check in appendResult before AppendFieldData
- Add comprehensive unit tests for empty and partial empty FieldsData
scenarios in both reduce and rerank functions
This fix handles both pure requery (all empty) and mixed scenarios (some
empty, some with data) without breaking normal search flow. The key
improvement is checking FieldsData length directly rather than IDs, as
requery may have IDs but empty FieldsData.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
Cherry-pick from master
pr: #45100
Related to #45081
StringIndexSort now properly handles collections with all-null string
fields by:
- Removing the error thrown when unique_count is 0 in ParseBinaryData
- Adding alignment and padding support in mmap serialization (similar to
ScalarIndexSort)
- Separating data_size_ from mmap_size_ to correctly parse data without
reading padding
This fixes load collection timeout failures when all string field data
is null, particularly affecting STL_SORT and TRIE index types.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
