https://github.com/milvus-io/milvus/issues/35528
This PR adds json index support for json and dynamic fields. Now you can
only do unary query like 'a["b"] > 1' using this index. We will support
more filter type later.
basic usage:
```
collection.create_index("json_field", {"index_type": "INVERTED",
"params": {"json_cast_type": DataType.STRING, "json_path":
'json_field["a"]["b"]'}})
```
There are some limits to use this index:
1. If a record does not have the json path you specify, it will be
ignored and there will not be an error.
2. If a value of the json path fails to be cast to the type you specify,
it will be ignored and there will not be an error.
3. A specific json path can have only one json index.
4. If you try to create more than one json indexes for one json field,
sdk(pymilvus<=2.4.7) may return immediately because of internal
implementation. This will be fixed in a later version.
---------
Signed-off-by: sunby <sunbingyi1992@gmail.com>
Related to #39840
The target could be updated async in previous code. This PR make remove
collection from target observer block until all tasks related in
dispatchers are removed preventing the metrics being updated after
collection released.
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #38399
- Make a timetick-commit-based write ahead buffer at write side.
- Add a switchable scanner at read side to transfer the state between
catchup and tailing read
Signed-off-by: chyezh <chyezh@outlook.com>
issue: #39680
if compaction/gc happens, load collection may stuck due to
SegmentNotFound, we should trigger UpdateNextTarget to get a new data
view to execute loading operation.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #39681
this PR maintain workload effect in action instead of computing workload
effect from target, which may cause leak if target changes.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #38399
- Add a pchannel level checkpoint for flush processing
- Refactor the recovery of flushers of wal
- make a shared wal scanner first, then make multi datasyncservice on it
Signed-off-by: chyezh <chyezh@outlook.com>
issue: #36621#39417
1. Adjust the server-side cache size.
2. Add source information for configurations.
3. Add node ID for compaction and indexing tasks.
4. Resolve localhost access issues to fix health check failures for
etcd.
Signed-off-by: jaime <yun.zhang@zilliz.com>
issue: #38399
- Embed the query node into streaming node to make delegator available
at streaming node.
- The embedded query node has a special server label
`QUERYNODE_STREAMING-EMBEDDED`.
- Change the balance strategy to make the channel assigned to streaming
node as much as possible.
Signed-off-by: chyezh <chyezh@outlook.com>
issue: https://github.com/milvus-io/milvus/issues/37630
Reduce the frequency updating metrics to avoid holding the mutex for
long periods.
---------
Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
When there are many segment tasks in the querycoord scheduler, the
traversal in `GetSegmentTaskDelta` checks becomes time-consuming. This
PR adds caching for segment deltas.
issue: https://github.com/milvus-io/milvus/issues/37630
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
Co-authored-by: Wei Liu <wei.liu@zilliz.com>
Read metadata such as segments, binlogs, and partitions concurrently at
the collection level.
issue: https://github.com/milvus-io/milvus/issues/37630
---------
Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
1. Provide partition&channel level indexing in the collection target.
2. Make `SegmentAction` not wait for distribution.
3. Remove scheduler and target manager mutex.
4. Optimize logging to reduce CPU overhead.
issue: https://github.com/milvus-io/milvus/issues/37630
---------
Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
issue: #38715
- Current milvus use a serialized index size(compressed) for estimate
resource for loading.
- Add a new field `MemSize` (before compressing) for index to estimate
resource.
---------
Signed-off-by: chyezh <chyezh@outlook.com>
issue: #38970
cause the stopping balance channel still use the row_count_based policy,
which may causes channel unbalance in multi-collection case.
This PR impl a score based stopping balance channel policy.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #39001
Background:
Segment Load Version: Each segment load request assigns a timestamp as
its version. When multiple copies of a segment are loaded on different
QueryNodes, the leader checker uses this version to identify the latest
copy and updates the routing table in the leader view to point to it.
Delegator Router Version: When a delegator builds a route to a QueryNode
that has loaded a segment, it also records the segment's version.
Router Table Update Logic: If the leader checker detects that the
version of a segment in the routing table does not match the version in
the worker, it updates the routing table to point to the QueryNode with
the latest version. Additionally, it updates the segment's load version
in the QueryNode during this process.
Issue:
When a channel is undergoing load balancing, the leader checker may sync
the routing table to a new delegator. This sync operation modifies the
segment's load version, which invalidates the routing in the old
delegator. Subsequently, the leader checker updates the routing table in
the old delegator, breaking the routing in the new delegator. This cycle
continues, causing repeated updates and inconsistencies.
Fix:
This PR introduces two changes to address the issue:
1. Use NodeID to verify whether the delegator's routing table needs an
update, avoiding unnecessary modifications.
2. Ensure compatibility by using the latest segment's load version as
the version recorded in the routing table.
These changes resolve the cyclic updates and prevent the leader checker
from generating excessive duplicate tasks, ensuring routing stability
across delegators during load balancing.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
fix#38649
when partition load failed, the partition drop will also fail due to the
wrong error message
Signed-off-by: xiaofanluan <xiaofan.luan@zilliz.com>
issue: #38718
The balancer calculates the workload of executing tasks as an ongoing
score for target nodes. However, a logic issue arises when
GetSegmentTaskDelta or GetChannelTaskDelta is called with
collectionID=-1, which incorrectly returns zero.
Due to the incorrect global score, the executing task's workload is not
properly reflected for each collection. Consequently, each collection
submits its own balance task, leading to the balancer assigning
excessive tasks to the same QueryNode.
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
1. A collection should observe the channel only once.
2. A collection should check the CollectionLoadPercent for updates only
once.
3. Skip saving coll/partition meta if there are no changes, primarily to
accelerate collection observation after recovery.
issue: https://github.com/milvus-io/milvus/issues/37630
Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
issue: #35563
1. Use an internal health checker to monitor the cluster's health state,
storing the latest state on the coordinator node. The CheckHealth
request retrieves the cluster's health from this latest state on the
proxy sides, which enhances cluster stability.
2. Each health check will assess all collections and channels, with
detailed failure messages temporarily saved in the latest state.
3. Use CheckHealth request instead of the heavy GetMetrics request on
the querynode and datanode
Signed-off-by: jaime <yun.zhang@zilliz.com>
the task limit in assignSegment/assignChannel will works for both load
task and balance task.
this PR remove the load task limit, only limit balance task num in one
round.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #38142
current balance channel policy only consider current collection's
distribution, so if all collections has 1 channel, and all channels has
been loaded on same querynode, after querynode num increase, balance
channel won't be triggered.
This PR enable score based balance channel policy, to achieve:
1. distribute all channels evenly across multiple querynodes
2. distribute each collection's channel evenly across multiple
querynodes.
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: https://github.com/milvus-io/milvus/issues/37764
- After removing rpc layer from mixcoord, the querycoord at standby mode
will be blocked forever of deployment rolling
---------
Signed-off-by: chyezh <chyezh@outlook.com>
issue: #38237
this PR only use better compression level for proto msg which is larger
than 1MB, and use a lighter compression level for smaller proto msg,
which could get a better latency in most case.
this PR could reduce the latency from 22.7s to 4.7s with 10000
collctions and each collections has 1000 segments.
before this PR:
BenchmarkTargetManager-8 1 22781536357 ns/op 566407275088 B/op 11188282
allocs/op
after this PR:
BenchmarkTargetManager-8 1 4729566944 ns/op 36713248864 B/op 10963615
allocs/op
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #38201
leader task require to update delegator's distribution, and only success
after the distribution change has been applyed to delegator. but the
delegator will reject the distribution change if it's version is older
than current version in delegator. which cause the leader task stuck and
retry forever.
this PR remove the leader task finish check.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #38305
after we disable balance segment and balance channel happens at same
time, the constriant which require release segment must happens on
serviceable shard leader is unnessary.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
Related to #36456
Unify collection/partition number metrics to collection manager in case
of unwant missing modification
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #35917
This PR refine the querycoord meta related interfaces to ensure that
each method includes a ctx parameter.
Signed-off-by: tinswzy <zhenyuan.wei@zilliz.com>
Related to #37630
This PR add a new util coll2Replicas secondary index to reduce map
access & iteration while get replicas by collection
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #33550
balance segment and balance segment execute at same time, which will
cause bounch of corner case.
This PR disable simultaneous balance of segments and channels
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #37830
casue dist handler doesn't set channel's version, so if channel checker
try to dedup channel, it may release the new delegator after balance
finished.
this PR fix the way to set proper version for channel.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #37679
pr #36549 introduce the logic error which update current target when
only parts of channel is ready.
This PR fix the logic error and let dist handler keep pull distribution
on querynode until all delegator becomes serviceable.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: ##36621
- For simple types in a struct, add "string" to the JSON tag for
automatic string conversion during JSON encoding.
- For complex types in a struct, replace "int64" with "string."
Signed-off-by: jaime <yun.zhang@zilliz.com>
issue: #37640
fix the pr #36549
cause balance channel will wait until new delegator becomes serviceable,
but new delegator need to sync target version then becomes serviceable,
and sync target version need to be wait all replica load done. so if
increasing replica number and balance channel happens at same time,
logic dead lock occurs.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #36293#36242
after qn recover, delegator may be loaded in new node, after all segment
has been loaded, delegator becomes serviceable. but delegator's target
version hasn't been synced, and if search/query comes, delegator will
use wrong target version to filter out a empty segment list, which
caused empty search result.
This pr will block delegator's serviceable status until target version
is synced
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
Related to #35415
In rolling upgrade, legacy proxy may dispatch load request wit empty
load field list. The upgraded querycoord may report error by mistake
that load field list is changed.
This PR:
- Auto field empty load field list with all user field ids
- Refine the error messag when load field list updates
- Refine load job unit test with service cases
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
When there're a lot of loaded collections, they would occupy the target
observer scheduler’s pool. This prevents loading collections from
updating the current target in time, slowing down the load process.
This PR adds a separate target dispatcher for loading collections.
issue: https://github.com/milvus-io/milvus/issues/37166
---------
Signed-off-by: bigsheeper <yihao.dai@zilliz.com>
When check health logic failed to collection not-queryable, the related
reason is hard to find in log.
This PR add context for log with trace id and print unqueryable
collection info log.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #36977
with node_label_filter on resource group, user can add label on
querynode with env `MILVUS_COMPONENT_LABEL`, then resource group will
prefer to accept node which match it's node_label_filter.
then querynode's can't be group by labels, and put querynodes with same
label to same resource groups.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #36970
cause release segment and balance channel may happen at same time, and
before new delegator become serviceable, if release segment exeuctes on
new delegator, and search/query comes on old delegator, then release
segment and query segment happens in parallel, if release segment
execute first in worker, then search/query will got a SegmentNodeLoaded
error.
This PR add serviceable filter on delegator, then all load/release
segment operation will happens on serviceable delegator.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
See also #37205
Previously releasing growing segments could be triggered by two
conditions:
- Sealed Segment with same id is loaded
- Segment start position is before target checkpoint ts
Which has a worst case that the corresponding sealed segment is
compacted and the checkpoint is pinned by a growing l0 segment.
This PR introduces a new rule that: a growing segment could be released
if the segment id appeared in current target dropped segment id list.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #36621
1. Add API to access task runtime metrics, including:
- build index task
- compaction task
- import task
- balance (including load/release of segments/channels and some leader
tasks on querycoord)
- sync task
2. Add a debug model to the webpage by using debug=true or debug=false
in the URL query parameters to enable or disable debug mode.
Signed-off-by: jaime <yun.zhang@zilliz.com>
issue: #37054
after querycoord restart, segment_checker may release segment by mistake
due to next target isn't ready yet.
This PR requires release segment must happens after next target is
ready.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
Relatedt #36887
DirectFoward streaming delete will cause memory usage explode if the
segments number was large. This PR add batching delete API and using it
for direct forward implementation.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #33550
cause wrong impl of UpdateCollectionNextTarget, if ReleaseCollection and
UpdateCollectionNextTarget happens at same time, the the released
partition's segment list may be add to target again, and delegator will
be marked as unserviceable due to lack of segment.
This PR fix the impl of UpdateCollectionNextTarget
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #36464
This PR enable balance on querynode with different mem capacity, for
query node which has more mem capactity will be assigned more records,
and query node with the largest difference between assignedScore and
currentScore will have a higher priority to carry the new segment.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #36488
when call TransferChannel/TransferSegment, querycoord will generate and
submit balance task to scheduler, if segment/channel's task already
exist in scheduler, submit task will failed.
to make TransferChannel/TransferSegment idempotent, we should skip to
submit if task already exist in scheduler.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #36536
query coord use `segmentTaskDeleta/channelTaskDelta` to measure the
executing workload for querynode in scheduler, and we maintains the
`segmentTaskDeleta/channelTaskDelta` by `scheulder.Add(task)` and
`scheduler.remove(task)`, but `scheduler.remove(task)` has been called
in unexpected way, which cause a wrong
`segmentTaskDeleta/channelTaskDelta` value and affect the segment assign
logic, causes segment unbalance.
This PR moves to compute the `segmentTaskDeleta/channelTaskDelta` when
access, to avoid the wrong value affect.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #35821
After collection loaded, if we need to increase/decrease collection's
replica, we need to release and load it again.
milvus offers 4 solution to update loaded collection's replica, this PR
aims to dynamic change the replica number without release, and after
replica number changed, milvus will execute load replica or release
replica in async, and the replica loaded status can be checked by
getReplicas API.
Notice that if set too much replicas than querynode can afford，the new
replica won't be loaded successfully until enough querynode joins.
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #36426
the old constriant requires only segment on current target can be
balanced, which is wrong, and caused that segment can't be move out from
stopping node, if it's only exist in next target.
by design, stopping balance need to move out all segment on it by
balance task, thus the unfair old constriant should be removed.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #36201
after querynode has been remove from replica, all dirty segment/channel
on it should be released.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
yihao.dai
wei liu
Zhen Ye
Bingyi Sun
congqixia
jaime
Xiaofan
SimFG
tinswzy
sthuang
aoiasd
cai.zhang
Jiquan Long