issue: https://github.com/milvus-io/milvus/issues/40006
This PR make tantivy document add by batch. Add document by batch can
greately reduce the latency of scheduling the document add operation
(call tantivy `add_document` only schdules the add operation and it
returns immediately after scheduled) , because each call involes a tokio
block_on which is relatively heavy.
Reduce scheduling part not necessarily reduces the overall latency if
the index writer threads does not process indexing quickly enough.
But if scheduling itself is pretty slow, even the index writer threads
process indexing very fast (by increasing thread number), the overall
performance can still be limited.
The following codes bench the PR (Note, the duration only counts for
scheduling without commit)
```
fn test_performance() {
let field_name = "text";
let dir = TempDir::new().unwrap();
let mut index_wrapper = IndexWriterWrapper::create_text_writer(
field_name,
dir.path().to_str().unwrap(),
"default",
"",
1,
50_000_000,
false,
TantivyIndexVersion::V7,
)
.unwrap();
let mut batch = vec![];
for i in 0..1_000_000 {
batch.push(format!("hello{:04}", i));
}
let batch_ref = batch.iter().map(|s| s.as_str()).collect::<Vec<_>>();
let now = std::time::Instant::now();
index_wrapper
.add_data_by_batch(&batch_ref, Some(0))
.unwrap();
let elapsed = now.elapsed();
println!("add_data_by_batch elapsed: {:?}", elapsed);
}
```
Latency roughly reduces from 1.4s to 558ms.
---------
Signed-off-by: SpadeA <tangchenjie1210@gmail.com>
issue: #40729
Current approach to parse negative numbers is first parse the numeric
part and then multiply the result by -1(mainly to distinguish the
precedence of the negative sign and the subtraction operator). However,
for the minimum value of int64(`-9223372036854775808`), the value
`9223372036854775808` already exceeds the representable range of int64.
As a result, parsing error occurs.
Therefore, use a specific rule to match `-9223372036854775808`.
---------
Signed-off-by: Cai Zhang <cai.zhang@zilliz.com>
Related to #41034
Recent pr #40842 introduced logic to avoid requery pk column, which
updates the original request which makes the request not equavilant to
the original one.
When retry happens due to incomplete request error, this change makes
the final result set lacks the pk column even when user specifies it
explicitly.
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
related to: #40884#39552
Reduce log frequency by aggregating deletion logs for L0 segments:
- Add segment count statistics in rangeHitL0Deletions function
- Change individual segment logs to a single consolidated log entry
- Include total number of processed L0 segments in log output
This change significantly reduces log volume while maintaining essential
visibility into deletion operations.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #41048
Fixes issue introduced in PR #33522 where metric resets caused
incomplete data collection by monitoring systems.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #39937
Before PR #39552, whenever a segment was missing in either the `current
target` or the `next target`, we would trigger `load segment` to recover
the delegator. However, restoring only the missing segments in the `next
target` is sufficient to advance the target and complete the recovery
process.
In PR #39552, we removed the scheduling of L0 segments along with this
unnecessary `load segment` logic. However, this exposed a new issue: if
the `current target` still has missing segments and there is a flaw in
the `checkDelegatorDataReady` logic, it could block the recovery of a
delegator that contains `offline segments`.
Since `offline segments` are cleaned up asynchronously in this scenario,
this PR removes their blocking effect on delegator recovery, ensuring a
smoother failure recovery process.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
The fields and partitions information are stored and fetched with
different prefixes in the metadata. In the CreateCollectionTask, the
RootCoord checks the existing collection information against the
metadata. This check fails if the order of the fields or partitions info
differs, leading to an error after restarting Milvus. To resolve this,
we should use a map in the check logic to ensure consistency.
related: https://github.com/milvus-io/milvus/issues/40955
---------
Signed-off-by: shaoting-huang <shaoting.huang@zilliz.com>
issue: #40638
- Add `ChannelID` for streaming replica in future.
- Remove the pchannel count fair balance policy for streaming.
- Add Score based vchannel fair balance policy for streaming.
- Add pchannel stats manager to collect the stats of pchannel for
balancer.
- Add configuration and metrics for new balance policy
---------
Signed-off-by: chyezh <chyezh@outlook.com>
fix: https://github.com/milvus-io/milvus/issues/40823
To solve the problem in the issue, we have to support building tantivy
index with low version
for those query nodes with low tantivy version.
This PR does two things:
1. refactor codes for IndexWriterWrapper to make it concise
2. enable IndexWriterWrapper to build tantivy index by different tantivy
crate
---------
Signed-off-by: SpadeA <tangchenjie1210@gmail.com>
after the pr merged, we can support to insert, upsert, build index,
query, search in the added field.
can only do the above operates in added field after add field request
complete, which is a sync operate.
compact will be supported in the next pr.
#39718
---------
Signed-off-by: lixinguo <xinguo.li@zilliz.com>
Co-authored-by: lixinguo <xinguo.li@zilliz.com>
`incomplete query result, missing id %!s(int64=348), len(searchIDs) =
10, len(queryIDs) = 9` error message format with error when missing id
is int64
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #40532
- balance should enable only when there's no proxy and datanode which
version is lower than 2.6.0
---------
Signed-off-by: chyezh <chyezh@outlook.com>
With concurrenct L0 compaction
(https://github.com/milvus-io/milvus/pull/36816), delta logs might be
written to the same L1 segment, causing logID duplication when using the
incremental beginLogID. This PR removes the beginLogID mechanism and
instead passes a log ID range, where the number of IDs in the range
equals the number of compaction segment binlogs multiplied by an
expansion factor.
issue: https://github.com/milvus-io/milvus/issues/40207
---------
Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
issue: #37651
this PR enable to balance the collection with largest row count first,
to avoid temporary migration of small table data to new nodes during
their onboarding, only to be moved out again after the large table
balance, which would cause unnecessary load.
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
proxy to always remove pk field from output field when forwarding
request to QN, and if user requested pk, fill it from IDs.
issue: https://github.com/milvus-io/milvus/issues/40833
---------
Signed-off-by: Buqian Zheng <zhengbuqian@gmail.com>
