8717 Commits

Author SHA1 Message Date
XuanYang-cn
efdbd8e7c1
enhance: Enable to upload by batch (#32788)
Signed-off-by: yangxuan <xuan.yang@zilliz.com>
2024-05-13 19:07:32 +08:00
XuanYang-cn
29b621f759
fix: Make compactor able to clear empty segments (#32821)
See also: #32553

Signed-off-by: yangxuan <xuan.yang@zilliz.com>
2024-05-13 18:21:32 +08:00
jaime
ba625835bc
enhance: Add metrics for segment index files size (#32979)
issue:#32980

Signed-off-by: jaime <yun.zhang@zilliz.com>
2024-05-13 17:59:32 +08:00
chyezh
293f14a8b9
fix: remove redundant replica recover (#32985)
issue: #22288 

- replica recover should be only triggered by replica recover

Signed-off-by: chyezh <chyezh@outlook.com>
2024-05-13 15:25:32 +08:00
foxspy
f6777267e3
enhance: add score compute consistency config for knowhere (#32997)
issue: https://github.com/milvus-io/milvus/issues/32583
related: #32584

Signed-off-by: xianliang.li <xianliang.li@zilliz.com>
2024-05-13 14:21:31 +08:00
congqixia
12ec3d61d9
fix: Fill deltalog entry num & time range in L0 compactions (#33004)
Resolves #33003

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2024-05-13 14:13:31 +08:00
Xiaofan
230c88b1de
enhance: improve reduce performance on very large topk (#32871)
fix #32870
reduce cpu usage for reduce on large k

Signed-off-by: xiaofanluan <xiaofan.luan@zilliz.com>
2024-05-13 11:55:31 +08:00
congqixia
0e5765b116
enhance: Utilize TestLocations ability to accelerate write & compaction (#32948)
See also #32642

This PR reuses hash locations for bloom filter prediction utilizing
`storage.Location`, like enhancement #32642.

Also adds a utility struct in storage: `LocationCache` to storage
locations for variable K (numbers of hash functions)

---------

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2024-05-13 10:15:32 +08:00
wei liu
8a9a42198d
fix: Proxy crash due to shard leader cache data race (#32971)
issue: #32970
cause InvalidateShardLeaderCache use wrong lock, which may cause data
race in meta cache, then proxy may crash

This PR fixed that use leaderMut when try to access shard leader cache.

Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2024-05-11 14:32:12 +08:00
Xiaofan
36f1ea93a5
enhance: optimize plan parser pool to avoid unnessary recycle (#32869)
fix #32868
plan parser takes too much cpu on high qps,this pr try to avoid create
lexer and parser too freequent

Signed-off-by: xiaofanluan <xiaofan.luan@zilliz.com>
2024-05-11 10:51:31 +08:00
Xiaofan
b044e5503e
enhance:Improve load speed (#32898)
fix #32897
add memory check when load collection

Signed-off-by: xiaofanluan <xiaofan.luan@zilliz.com>
2024-05-11 10:29:31 +08:00
congqixia
c0e62e6629
enhance: Use collection default consistency level for restv2 (#32956)
Set `UseDefaultConsistency` to true so that restv2 read API shall use
collection consistency level setting correctly.

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2024-05-11 00:47:31 +08:00
chyezh
1c84a1c9b6
fix: lru related issue fixup patch (#32916)
issue: #32206, #32801

- search failure with some assertion, segment not loaded and resource
insufficient.

- segment leak when query segments

---------

Signed-off-by: chyezh <chyezh@outlook.com>
2024-05-10 19:17:30 +08:00
wei liu
25689859a1
fix: Load index metric use wrong time unit (#32935)
issue:#32899

This PR fix the wrong metric value of load index, which introduced by
pr#32567, use wrong time unit for load index metrics

Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2024-05-10 18:07:30 +08:00
wei liu
e2332bdc17
enhance: Enable channel exclusive balance policy (#32911)
issue: #32910  
* split replica's node list to channels when create replicas
 * balance nodes among channels when node change happens
 * implement channel level balance, let balance happens in channel level

Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2024-05-10 17:27:31 +08:00
Gao
f7d29118a5
enhance: update knowhere version (#32931)
Signed-off-by: chasingegg <chao.gao@zilliz.com>
2024-05-10 16:13:31 +08:00
Cai Yudong
dc89c6f810
enhance: remove duplicated data generation APIs for bulk insert test (#32889)
Issue: #22837

including following changes:
1. Add API CreateInsertData() and BuildArrayData() in
internal/util/testutil
2. Remove duplicated test APIs from importutilv2 unittest and bulk
insert integration test

Signed-off-by: Cai Yudong <yudong.cai@zilliz.com>
2024-05-10 15:27:31 +08:00
wei liu
04a8ec69f6
fix: Segment on stopping query node can't be release successfully (#32929)
issue: #32901
Cause release segment request need be send to delegator, but it need
replica to info find segment's delegator. but the stopping query node
will be marked as read only in replica, then `replica.Contains()` just
return true for rwNode in replica. then it can't get replica info by
stopping query node and release segment will be blocked.

This PR make `replica.Contains()` return true for both roNode and
rwNode.

Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2024-05-10 14:33:30 +08:00
Jiquan Long
0783582e2e
fix: temporarily disable two-phase retrieval when lru is enabled (#32927)
issue: #31822

Signed-off-by: longjiquan <jiquan.long@zilliz.com>
2024-05-10 14:19:45 +08:00
Bingyi Sun
5cbf081111
fix: fix index resource estimation (#32842)
issue: #32820

---------

Signed-off-by: sunby <sunbingyi1992@gmail.com>
2024-05-10 11:53:30 +08:00
Bingyi Sun
b7ef8da360
fix: set channel checkpoint to delta position (#32878)
issue: https://github.com/milvus-io/milvus/issues/32853

---------

Signed-off-by: sunby <sunbingyi1992@gmail.com>
2024-05-10 11:51:30 +08:00
Buqian Zheng
bb7765cbd6
fix: fix Indexing.Iterator ut: build index with all data at once (#32844)
issue: #32843

Signed-off-by: Buqian Zheng <zhengbuqian@gmail.com>
2024-05-10 11:31:30 +08:00
wei liu
c35797c399
enhance: expose DescribeDatabase api in proxy (#32732)
issue: #32707

---------

Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2024-05-09 22:51:30 +08:00
congqixia
efa58ae423
enhance: Utilize coll2replica mapping when getting rg by collection (#32892)
See also #32165

In old `GetResourceGroupByCollection` implementation, it iterates all
replicas to match collection id, which is slow and CPU time consuming.
This PR make it utilize the coll2Replicas mapping by calling
`GetByCollection` and mapping replicas into resource group.

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2024-05-09 19:37:30 +08:00
congqixia
77fa615772
fix: Make SyncManager callback func ignore nil error (#32891)
introduced by #32865

sync manager callback handler panicked when error is nil

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2024-05-09 18:03:31 +08:00
aoiasd
54a51b1236
enhance: Support dynamic config for opentelemetry trace (#32169)
relate: https://github.com/milvus-io/milvus/issues/31940

---------

Signed-off-by: aoiasd <zhicheng.yue@zilliz.com>
2024-05-09 17:43:30 +08:00
congqixia
acb0417a9f
enhance: Avoid iteration over channel results when update leaderview (#32887)
See also #32165

Cache channel name to channel info to avoid iteration over channel
results when updating leader view version.

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2024-05-09 15:41:30 +08:00
XuanYang-cn
bc015df962
fix: [skip-e2e] Fix unstable policy_test (#32883)
See also: #32849

Signed-off-by: yangxuan <xuan.yang@zilliz.com>
2024-05-09 15:39:40 +08:00
Cai Yudong
8bb58d0460
enhance: optimize vector offsets handling for parquet (#32822)
Issue: #22837

Signed-off-by: Cai Yudong <yudong.cai@zilliz.com>
2024-05-09 14:43:30 +08:00
congqixia
a06f601c6e
fix: Make syncmgr lock key before returning future (#32865)
See also #32860

SyncMgr did not ensure task key is locked before `SyncData` returning
which may cause concurrent problem during sync wich multiple policies.

This PR change sync mgr implementation to make sure the key is locked
before returning task result `*conc.Future`

---------

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2024-05-09 10:09:30 +08:00
sre-ci-robot
3d78b90fe7
[automated] Update Knowhere Commit (#32872)
Update Knowhere Commit
Signed-off-by: sre-ci-robot sre-ci-robot@users.noreply.github.com

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
2024-05-09 01:57:30 +08:00
congqixia
cedb33ceec
enhance: Improve datacoord segment filtering with collection (#32831)
See also #32165

This PR modify the `SelectSegments` interface to utilizing collection id
information when selecting segment with provided collection

---------

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2024-05-08 21:37:29 +08:00
Jiquan Long
035a508722
fix: make sure inverted index has only one segment (#32858)
issue: #32717

---------

Signed-off-by: longjiquan <jiquan.long@zilliz.com>
2024-05-08 21:25:30 +08:00
XuanYang-cn
0b0d279515
fix: [skip-e2e] Fix AvgAssignByCount unstable (#32850)
See also: #32849

Signed-off-by: yangxuan <xuan.yang@zilliz.com>
2024-05-08 19:09:35 +08:00
Bingyi Sun
17a79f4ca9
enhance: The LRU cache evicts items and retries loading if the disk limit is reached. (#32819)
Signed-off-by: sunby <sunbingyi1992@gmail.com>
2024-05-08 14:45:30 +08:00
yihao.dai
a8db16ad4b
enhance: Change ddl log to info (#32838)
/kind improvement

Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
2024-05-08 14:17:50 +08:00
yiwangdr
d6e537c91c
fix: allow datanode's server id to be updated (#31597)
issue: #31516

background: the server id field in data node is redundant. session id
already provides the source of truth.

Signed-off-by: yiwangdr <yiwangdr@gmail.com>
2024-05-08 14:03:29 +08:00
Chun Han
01c2684355
enhance: [skip e2e] disable unstable ut temporarily (#32836)
Signed-off-by: MrPresent-Han <chun.han@zilliz.com>
2024-05-08 12:17:29 +08:00
SimFG
8963615b33
enhance: add the cost response for the rest api (#32620)
issue: https://github.com/milvus-io/milvus/issues/30436

Signed-off-by: SimFG <bang.fu@zilliz.com>
2024-05-08 11:53:29 +08:00
wei liu
fad8f0afa5
enhance: enable stopping balance after balance has been suspended (#32812)
issue: #32811

Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2024-05-08 10:15:29 +08:00
wei liu
ba02d54a30
enhance: update shard leader cache when leader location changed (#32470)
issue: #32466

this PR enhance that when shard location changed, update proxy's shard
leader cache. in case of query node failover case, proxy can find
replica recover

---------

Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2024-05-08 10:05:29 +08:00
wei liu
5038036ece
enhance: Reuse hash locations during access bloom fitler (#32642)
issue: #32530 

when try to match segment bloom filter with pk, we can reuse the hash
locations. This PR maintain the max hash Func, and compute hash location
once for all segment, reuse hash location can speed up bf access

---------

Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2024-05-07 06:13:47 -07:00
Jiquan Long
9837ad6a8d
enhance: remove deprecated api (#32808)
issue: #32728

Signed-off-by: longjiquan <jiquan.long@zilliz.com>
2024-05-07 20:03:30 +08:00
congqixia
40728ce83d
enhance: Add metautil.Channel to convert string compare to int (#32749)
See also #32748

This PR:

- Add `metautil.Channel` utiltiy which convert virtual name to physical
channel name, collectionID and shard idx
- Add channel mapper interface & implementation to convert limited
physical channel name into int index
- Apply `metautil.Channel` filter in querynode segment manager logic

---------

Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
2024-05-07 19:13:35 +08:00
XuanYang-cn
6843d6d376
fix: Compaction trigger choose 2 same segments (#32800)
DataNode would stuck at compactor try to lock the
same segmentID

See also: #32765

---------

Signed-off-by: yangxuan <xuan.yang@zilliz.com>
2024-05-07 19:01:31 +08:00
Cai Yudong
bcdbd1966e
feat: Support sparse float vector bulk insert for binlog/json/parquet (#32649)
Issue: #22837

Signed-off-by: Cai Yudong <yudong.cai@zilliz.com>
2024-05-07 18:43:30 +08:00
yihao.dai
53874ce245
fix: Fix cannot specify partition name in binlog import (#32730)
issue: https://github.com/milvus-io/milvus/issues/32807

---------

Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
2024-05-07 17:19:30 +08:00
yihao.dai
9db3aa18bc
enhance: Remove deprecated EnableIndex (#32704)
/kind improvement

Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
2024-05-07 17:11:30 +08:00
chyezh
641f702f64
fix: add request resource timeout for lazy load, refactor context usage in cache (#32709)
issue: #32663

- Use new param to control request resource timeout for lazy load.

- Remove the timeout parameter of `Do`, remove `DoWait`. use `context`
to control the timeout.

- Use `VersionedNotifier` to avoid notify event lost and broadcast,
remove the redundant goroutine in cache.

related dev pr: #32684

Signed-off-by: chyezh <chyezh@outlook.com>
2024-05-07 16:33:30 +08:00
yiwangdr
b1eacb2ae8
feat: datacoord/node watch based on rpc (#32036)
issue: https://github.com/milvus-io/milvus/issues/25309

Signed-off-by: yiwangdr <yiwangdr@gmail.com>
2024-05-07 15:49:30 +08:00