milvus-io
/
milvus
mirror of https://github.com/milvus-io/milvus.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
milvus/CHANGELOG.md

29 KiB
Raw Blame History

Changelog

Please mark all change in change log and use the issue from GitHub

Milvus 0.8.0 (TBD)

Bug

  • Server is not forbidden to create new partition which tag is "_default"

Feature

  • GPU index support delete vectors
  • Add annoy index type in C++ sdk

Improvement

Task

Milvus 0.7.1 (2020-03-29)

Bug

  • Data in WAL may be accidentally inserted into a new table with the same name.
  • Fix search demo bug in HTTP doc
  • Vectors can be returned by searching after vectors deleted if cache_insert_data set true
  • The cache cannot be used all when the vector type is binary
  • Check validity of dimension when collection metric type is binary one
  • PQ index parameter 'm' validation
  • API search_in_files cannot work correctly when vectors is stored in certain non-default partition
  • Fix SQ8H search fail on SIFT-1B dataset
  • Create index failed with type: rnsg if metric_type is IP
  • NSG search crashed
  • Remove unused unittests
  • Optimize request handler to combine similar query
  • Opentracing for combined search request
  • Fix search out of memory with ivf_flat
  • Expected error status if search with partition_tag not existed
  • Fix memory exhausted during searching
  • Fix search hang with SQ8H

Feature

  • Integrate ANNOY into Milvus
  • BinaryFlat add 2 Metric: Substructure and Superstructure
  • IVF PQ CPU support deleted vectors searching
  • HNSW support deleted vectors searching

Improvement

  • Improve search performance: reduce delay
  • Knowhere and Wrapper refactor
  • Optimize raw vector and uids read/write
  • Move Config.cpp to config directory
  • Rename storage/file to storage/disk and rename classes
  • Move store/Directory to storage/Operation and add FSHandler
  • Optimize config cpu/gpu cache_capacity setter
  • Improve compact performance
  • Fix Milvus crash on old CPU
  • IndexFlat (SSE) and IndexBinaryFlat performance improvement for small NQ
  • Remove CUSTOMIZATION macro
  • Upgrade mishards to v0.7.0
  • Improve Milvus log
  • Optimize behavior to get file ids from metadata in mishards
  • Add Substructure and Superstructure in http module

Task

Milvus 0.7.0 (2020-03-11)

Bug

  • Milvus crash when searching and building index simultaneously using SQ8H
  • Don't return partition table for show_tables
  • Server unittest run failed on low-end server
  • IVFTest.gpu_seal_test unittest failed
  • Judge branch error in CommonUtil.cpp
  • Server crash when create tables concurrently
  • Check gpu resources setting when assign repeated value
  • Table count set to 0 if no tables found
  • Improve error message when offset or page_size is equal 0
  • Check if partition name is valid
  • Check if table exists when show partitions
  • Check if table exists when try to delete partition
  • Optimize http insert and search speed
  • Check if partition name is legal
  • Check if table exists when show partitions
  • Check if table exists when try to delete partition
  • Optimize http insert and search speed
  • Add binary vectors support in http server
  • Improve error message when page size or offset is illegal
  • Check page_size or offset value to avoid float
  • Http server support load table into memory
  • Error log output continuously after server start
  • Server down caused by searching with index_type: HNSW
  • Update license declaration
  • Unit test failed when on CPU2GPU case
  • Negative distance value returned when searching with HNSW index type
  • Server crashed when searching vectors with GPU
  • Fix vectors results bug when getting vectors from segments
  • Index type changed to IDMAP after compacted
  • Server crashed during adding vectors
  • Fix duplicated ID number issue
  • Avoid possible race condition between delete and search
  • set_config for insert_buffer_size is wrong
  • Add set interfaces for WAL configurations
  • Fix big integer cannot pass to server correctly
  • Result is not correct when search vectors in multi partition, index type is RNSG
  • Table count did not match after deleting vectors and compact
  • Make cache_insert_data take effect in-service
  • Add setter API for config preload_table
  • Fix server crash when cache_insert_data enabled
  • Set table file with correct engine type in meta
  • Search with ivf_flat failed with open-dataset: sift-256-hamming
  • Degradation searching performance with metric_type: binary_idmap
  • Fix server/wal config setting bug
  • Index file not created after table and index created
  • Search crashed with Super-high dimensional binary vector
  • Too low recall for glove-200-angular, ivf_pq index
  • Meta engine type become IDMAP after dropping index for BINARY table
  • Set all existing bitset in cache when applying deletes
  • Row count incorrect if delete vectors then create index
  • Old segment folder not removed after merge/compact if create_index is called before adding data
  • Server down caused by failure to write file during concurrent mixed operations
  • Server down during mixed operations
  • External link bug in HTTP doc
  • Refine Compact function
  • Building index params check for Annoy

Feature

  • Add CLI to get server info
  • Add Opentracing
  • Support get/set config via CLI
  • Put C++ sdk out of milvus/core
  • If partition tag is similar, wrong partition is searched
  • Add server build commit info interface
  • Add web server into server module
  • Add push mode for prometheus monitor
  • Support MinIO storage
  • Support binary vector tanimoto/jaccard/hamming metric
  • Support WAL(write-ahead logging)
  • Support HNSW
  • Support DeleteById / SearchByID / GetVectorById / Flush
  • Change Milvus c++ standard to c++17
  • Support AVX-512 in FAISS
  • Add api to get table data information
  • Support CPU profiling
  • Get all record IDs in a segment by given a segment id
  • Add crud APIs and segments APIs into http module
  • Update config version to 0.2
  • Remove S3 related config

Improvement

  • Use Openblas / lapack from apt install
  • Enhance config description
  • Remove Arrow
  • Add cpu mode for built-in Faiss
  • Add ready-to-use config files to the Milvus repo for enhanced user experince
  • Remove redundant checks in CacheMgr's constructor
  • Move "primary_path" and "secondary_path" to storage config
  • Remove "collector" from config
  • Update NOTICE.md
  • Rename minio to s3 in Storage Config section
  • Move 'insert_buffer_size' to Cache Config section
  • Error message is not clear when creating IVFSQ8H index without gpu resources
  • , , , , , , , , , , , , , , , , , , , Various fixes and improvements for Milvus documentation.
  • Hide partition_name parameter, avid user directly access partition table
  • Do S3 server validation check when Milvus startup
  • Allow system conf modifiable and some take effect directly
  • Add default partition tag for a table
  • Remove debug logging from faiss
  • Support to configure whether to enabled autoflush and the autoflush interval
  • Improve delete
  • General proto api for NNS libraries
  • Add return code for AVX512 selection
  • Update config "preload_table" description
  • Update resources name in HTTP module
  • Update yaml config description

Task

  • Exclude third-party code from codebeat
  • Exclude third-party code from codacy

Milvus 0.6.0 (2019-12-07)

Bug

  • Memory usage increased slowly during searching vectors
  • Exclude src/external folder from code coverage for jenkin ci
  • Reside src/external in thirdparty
  • Some files not merged after vectors added
  • Search does not use GPU when index type is FLAT
  • Add exception handle when search fail
  • Test cases run failed on 0.6.0
  • Rename config.h.in to version.h.in
  • sdk_simple return empty result
  • Create partition success if tag name only contains spaces
  • sdk_simple return incorrect result
  • Create partition should be failed if partition tag existed
  • Message returned is confused when partition created with null partition name
  • Drop the same partition success repeatally
  • Query API in customization still uses old version
  • Server cannot startup with gpu_resource_config.enable=false in GPU version
  • Index data is not compatible between 0.5 and 0.6
  • Server hang caused by searching with nsg index
  • Increase code coverage rate
  • gpu no usage during index building
  • CPU-version search performance decreased
  • The code coverage rate of core/src/scheduler/optimizer is too low
  • IVF_PQ index build trapped into dead loop caused by invalid params
  • Unittest DELETE_BY_RANGE sometimes failed
  • Erase file data from cache once the file is marked as deleted
  • faiss benchmark not compatible with faiss 1.6.0
  • BuildIndex stop when do build index and search simultaneously
  • Assigin value to table_name from confest shell
  • NSG build failed with MetricType Inner Product
  • client raise exception in shards when search results is empty
  • Avoid dead circle of build index thread when error occurs
  • NSG build failed using GPU-edition if set gpu_enable false
  • NSG search accuracy is too low
  • Server down during building index_type: IVF_PQ using GPU-edition
  • Milvus server should report exception/error message or terminate on mysql metadata backend error
  • Build index hang in GPU version when gpu_resources disabled
  • Frequently insert operation cost too much disk space
  • Build index log is incorrect
  • Optimizer specify wrong gpu_id
  • No log generated during building index with CPU
  • IP search metric_type is not supported by IVF_PQ index
  • FAISS isn't compiled with O3 option
  • (CPU) Create index PQ should be failed if table metric type set Inner Product
  • Typo "partiton" should be "partition"
  • Random crash when frequently insert vector one by one
  • Milvus error out when building SQ8H index without GPU resources
  • Update badge of README
  • Random failure of unittest db_test::SEARCH_TEST
  • Server down in stability test
  • Metric_type changed from IP to L2
  • Fix search SQ8H crash without GPU resource

Feature

  • Pure CPU version for Milvus
  • Support table partition
  • Support new Index type IVFPQ
  • Experimental shards middleware for Milvus
  • Support new index types SPTAG-KDT and SPTAG-BKT
  • Support build index with multiple gpu
  • Update shards merge part to match v0.5.3
  • Add log in scheduler/optimizer
  • C++ SDK support IVFPQ and SPTAG
  • Add version in server config file
  • Print more messages when server start
  • Add a new rpc command to get milvus build version whether cpu or gpu
  • Show last commit id when server start

Improvement

  • Add ivfsq8 test report detailed version
  • C++ SDK README
  • Rpc request source code refactor
  • Logger the time cost during preloading data
  • Rename C++ SDK IndexType
  • Change C++ SDK to shared library
  • Use int64 for all config integer
  • Add Q&A for 'protocol https not supported or disable in libcurl' issue
  • add Find FAISS in CMake
  • Add option to enable / disable prometheus
  • Build migration scripts into milvus docker image
  • Add more information in build.sh and install.md
  • Add virtual method Init() in Pass abstract class
  • Add a Fallback pass in optimizer
  • C++ SDK query result is not easy to use
  • Add ShowPartitions example for C++ SDK
  • Small raw files should not be build index
  • Intergrate internal FAISS
  • Remove MILVUS_CPU_VERSION
  • FAISS GPU version is compiled with O0
  • Refactor server module to separate Grpc from server handler and scheduler

Task

Milvus 0.5.3 (2019-11-13)

Bug

  • Bytes type in proto cause big-endian/little-endian problem

Feature

Improvement

  • improve grpc performance in search
  • Add more unittest for config set/get
  • optimize unittest to support run single test more easily
  • Change C++ SDK to shared library
  • C++ SDK README

Task

Milvus 0.5.2 (2019-11-07)

Bug

  • Search faild: message="Table file doesn't exist"

Feature

Improvement

  • Update default config:use_blas_threshold to 1100 and server version printout to 0.5.2

Task

Milvus 0.5.1 (2019-11-04)

Bug

  • JFrog cache error
  • Search IVFSQHybrid crash on gpu
  • IVF_FLAT search out of memory

Feature

  • The server start error messages could be improved to enhance user experience
  • test_scheduler core dump
  • Using new structure for tasktable
  • New config option use_gpu_threshold
  • Add only GPU and only CPU version for IVF_SQ8 and IVF_FLAT
  • Add CPU version for building index

Improvement

  • Improvement dump function in scheduler
  • Print version information into log during server start
  • Move easyloggingpp into "external" directory
  • Speed up CMake build process
  • Remove .a file in milvus/lib for docker-version
  • Using shared_ptr instead of weak_ptr to avoid performance loss
  • Add unique id for Job
  • Set task state MOVED after resource copy it completed
  • Improve large query optimizer pass
  • Not return error when search_resources and index_build_device set cpu
  • Change the configuration name from 'use_gpu_threshold' to 'gpu_search_threshold'
  • Improve result reduce
  • add invalid config unittest

Task

Milvus 0.5.0 (2019-10-21)

Bug

  • MS-568 Fix gpuresource free error
  • MS-572 Milvus crash when get SIGINT
  • MS-577 Unittest Query randomly hung
  • MS-587 Count get wrong result after adding vectors and index built immediately
  • MS-599 Search wrong result when table created with metric_type: IP
  • MS-601 Docker logs error caused by get CPUTemperature error
  • MS-605 Server going down during searching vectors
  • MS-620 Get table row counts display wrong error code
  • MS-622 Delete vectors should be failed if date range is invalid
  • MS-624 Search vectors failed if time ranges long enough
  • MS-637 Out of memory when load too many tasks
  • MS-639 SQ8H index created failed and server hang
  • MS-640 Cache object size calculate incorrect
  • MS-641 Segment fault(signal 11) in PickToLoad
  • MS-644 Search crashed with index-type: flat
  • MS-647 grafana display average cpu-temp
  • MS-652 IVFSQH quantization double free
  • MS-650 SQ8H index create issue
  • MS-653 When config check fail, Milvus close without message
  • MS-654 Describe index timeout when building index
  • MS-658 Fix SQ8 Hybrid can't search
  • MS-665 IVF_SQ8H search crash when no GPU resource in search_resources
  • Change default gpu_cache_capacity to 4
  • C++ sdk example get grpc error
  • Add unittest to improve code coverage
  • make clang-format failed after run build.sh -l
  • Create SQ8H index hang if using github server version
  • Some troubleshoot messages in Milvus do not provide enough information
  • Config unittest failed
  • Topk result is incorrect for small dataset

Improvement

  • MS-552 Add and change the easylogging library
  • MS-553 Refine cache code
  • MS-555 Remove old scheduler
  • MS-556 Add Job Definition in Scheduler
  • MS-557 Merge Log.h
  • MS-558 Refine status code
  • MS-562 Add JobMgr and TaskCreator in Scheduler
  • MS-566 Refactor cmake
  • MS-574 Milvus configuration refactor
  • MS-578 Make sure milvus5.0 don't crack 0.3.1 data
  • MS-585 Update namespace in scheduler
  • MS-606 Speed up result reduce
  • MS-608 Update TODO names
  • MS-609 Update task construct function
  • MS-611 Add resources validity check in ResourceMgr
  • MS-619 Add optimizer class in scheduler
  • MS-626 Refactor DataObj to support cache any type data
  • MS-648 Improve unittest
  • MS-655 Upgrade SPTAG
  • Put union of index_build_device and search resources to gpu_pool
  • Avoid linking targets multiple times in cmake

Feature

  • MS-614 Preload table at startup
  • MS-627 Integrate new index: IVFSQHybrid
  • MS-631 IVFSQ8H Index support
  • MS-636 Add optimizer in scheduler for FAISS_IVFSQ8H

Task

  • MS-554 Change license to Apache 2.0
  • MS-561 Add contributing guidelines, code of conduct and README docs
  • MS-567 Add NOTICE.md
  • MS-569 Complete the NOTICE.md
  • MS-575 Add Clang-format & Clang-tidy & Cpplint
  • MS-586 Remove BUILD_FAISS_WITH_MKL option
  • MS-590 Refine cmake code to support cpplint
  • MS-600 Reconstruct unittest code
  • MS-602 Remove zilliz namespace
  • MS-610 Change error code base value from hex to decimal
  • MS-624 Re-organize project directory for open-source
  • MS-635 Add compile option to support customized faiss
  • MS-660 add ubuntu_build_deps.sh
  • Add all test cases

Milvus 0.4.0 (2019-09-12)

Bug

  • MS-119 The problem of combining the log files
  • MS-121 The problem that user can't change the time zone
  • MS-411 Fix metric unittest linking error
  • MS-412 Fix gpu cache logical error
  • MS-416 ExecutionEngineImpl::GpuCache has not return value cause crash
  • MS-417 YAML sequence load disable cause scheduler startup failed
  • MS-413 Create index failed and server exited
  • MS-427 Describe index error after drop index
  • MS-432 Search vectors params nprobe need to check max number
  • MS-431 Search vectors params nprobe: 0/-1, expected result: raise exception
  • MS-331 Crate Table : when table exists, error code is META_FAILED(code=15) rather than ILLEGAL TABLE NAME(code=9))
  • MS-430 Search no result if index created with FLAT
  • MS-443 Create index hang again
  • MS-436 Delete vectors failed if index created with index_type: IVF_FLAT/IVF_SQ8
  • MS-449 Add vectors twice success, once with ids, the other no ids
  • MS-450 server hang after run stop_server.sh
  • MS-458 Keep building index for one file when no gpu resource
  • MS-461 Mysql meta unittest failed
  • MS-462 Run milvus server twices, should display error
  • MS-463 Search timeout
  • MS-467 mysql db test failed
  • MS-470 Drop index success, which table not created
  • MS-471 code coverage run failed
  • MS-492 Drop index failed if index have been created with index_type: FLAT
  • MS-493 Knowhere unittest crash
  • MS-453 GPU search error when nprobe set more than 1024
  • MS-474 Create index hang if use branch-0.3.1 server config
  • MS-510 unittest out of memory and crashed
  • MS-507 Dataset 10m-512, index type sq8，performance in-normal when set CPU_CACHE to 16 or 64
  • MS-543 SearchTask fail without exception
  • MS-582 grafana displays changes frequently

Improvement

  • MS-327 Clean code for milvus
  • MS-336 Scheduler interface
  • MS-344 Add TaskTable Test
  • MS-345 Add Node Test
  • MS-346 Add some implementation of scheduler to solve compile error
  • MS-348 Add ResourceFactory Test
  • MS-350 Remove knowhere submodule
  • MS-354 Add task class and interface in scheduler
  • MS-355 Add copy interface in ExcutionEngine
  • MS-357 Add minimum schedule function
  • MS-359 Add cost test in new scheduler
  • MS-361 Add event in resource
  • MS-364 Modify tasktableitem in tasktable
  • MS-365 Use tasktableitemptr instead in event
  • MS-366 Implement TaskTable
  • MS-368 Implement cost.cpp
  • MS-371 Add TaskTableUpdatedEvent
  • MS-373 Add resource test
  • MS-374 Add action definition
  • MS-375 Add Dump implementation for Event
  • MS-376 Add loader and executor enable flag in Resource avoid diskresource execute task
  • MS-377 Improve process thread trigger in ResourceMgr, Scheduler and TaskTable
  • MS-378 Debug and Update normal_test in scheduler unittest
  • MS-379 Add Dump implementation in Resource
  • MS-380 Update resource loader and executor, work util all finished
  • MS-383 Modify condition variable usage in scheduler
  • MS-384 Add global instance of ResourceMgr and Scheduler
  • MS-389 Add clone interface in Task
  • MS-390 Update resource construct function
  • MS-391 Add PushTaskToNeighbourHasExecutor action
  • MS-394 Update scheduler unittest
  • MS-400 Add timestamp record in task state change function
  • MS-402 Add dump implementation for TaskTableItem
  • MS-406 Add table flag for meta
  • MS-403 Add GpuCacheMgr
  • MS-404 Release index after search task done avoid memory increment continues
  • MS-405 Add delete task support
  • MS-407 Reconstruct MetricsCollector
  • MS-408 Add device_id in resource construct function
  • MS-409 Using new scheduler
  • MS-413 Remove thrift dependency
  • MS-410 Add resource config comment
  • MS-414 Add TaskType in Scheduler::Task
  • MS-415 Add command tasktable to dump all tasktables
  • MS-418 Update server_config.template file, set CPU compute only default
  • MS-419 Move index_file_size from IndexParam to TableSchema
  • MS-421 Add TaskLabel in scheduler
  • MS-422 Support DeleteTask in Multi-GpuResource case
  • MS-428 Add PushTaskByDataLocality in scheduler
  • MS-440 Add DumpTaskTables in sdk
  • MS-442 Merge Knowhere
  • MS-445 Rename CopyCompleted to LoadCompleted
  • MS-451 Update server_config.template file, set GPU compute default
  • MS-455 Distribute tasks by minimal cost in scheduler
  • MS-460 Put transport speed as weight when choosing neighbour to execute task
  • MS-459 Add cache for pick function in tasktable
  • MS-476 Improve search performance
  • MS-482 Change search stream transport to unary in grpc
  • MS-487 Define metric type in CreateTable
  • MS-488 Improve code format in scheduler
  • MS-495 cmake: integrated knowhere
  • MS-496 Change the top_k limitation from 1024 to 2048
  • MS-502 Update tasktable_test in scheduler
  • MS-504 Update node_test in scheduler
  • MS-505 Install core unit test and add to coverage
  • MS-508 Update normal_test in scheduler
  • MS-532 Add grpc server unittest
  • MS-511 Update resource_test in scheduler
  • MS-517 Update resource_mgr_test in scheduler
  • MS-518 Add schedinst_test in scheduler
  • MS-519 Add event_test in scheduler
  • MS-520 Update resource_test in scheduler
  • MS-524 Add some unittest in event_test and resource_test
  • MS-525 Disable parallel reduce in SearchTask
  • MS-527 Update scheduler_test and enable it
  • MS-528 Hide some config used future
  • MS-530 Add unittest for SearchTask->Load
  • MS-531 Disable next version code
  • MS-533 Update resource_test to cover dump function
  • MS-523 Config file validation
  • MS-539 Remove old task code
  • MS-546 Add simple mode resource_config
  • MS-570 Add prometheus docker-compose file
  • MS-576 Scheduler refactor
  • MS-592 Change showtables stream transport to unary

Feature

  • MS-343 Implement ResourceMgr
  • MS-338 NewAPI: refine code to support CreateIndex
  • MS-339 NewAPI: refine code to support DropIndex
  • MS-340 NewAPI: implement DescribeIndex

Task

  • MS-297 disable mysql unit test

Milvus 0.3.1 (2019-07-10)

Bug

  • MS-148 Disable cleanup if mode is read only
  • MS-149 Fixed searching only one index file issue in distributed mode
  • MS-153 Fix c_str error when connecting to MySQL
  • MS-157 Fix changelog
  • MS-190 Use env variable to switch mem manager and fix cmake
  • MS-217 Fix SQ8 row count bug
  • MS-224 Return AlreadyExist status in MySQLMetaImpl::CreateTable if table already exists
  • MS-232 Add MySQLMetaImpl::UpdateTableFilesToIndex and set maximum_memory to default if config value = 0
  • MS-233 Remove mem manager log
  • MS-230 Change parameter name: Maximum_memory to insert_buffer_size
  • MS-234 Some case cause background merge thread stop
  • MS-235 Some test cases random fail
  • MS-236 Add MySQLMetaImpl::HasNonIndexFiles
  • MS-257 Update bzip2 download url
  • MS-288 Update compile scripts
  • MS-330 Stability test failed caused by server core dumped
  • MS-347 Build index hangs again
  • MS-382 fix MySQLMetaImpl::CleanUpFilesWithTTL unknown column bug

Improvement

  • MS-156 Add unittest for merge result functions
  • MS-152 Delete assert in MySQLMetaImpl and change MySQLConnectionPool impl
  • MS-204 Support multi db_path
  • MS-206 Support SQ8 index type
  • MS-208 Add buildinde interface for C++ SDK
  • MS-212 Support Inner product metric type
  • MS-241 Build Faiss with MKL if using Intel CPU; else build with OpenBlas
  • MS-242 Clean up cmake and change MAKE_BUILD_ARGS to be user defined variable
  • MS-245 Improve search result transfer performance
  • MS-248 Support AddVector/SearchVector profiling
  • MS-256 Add more cache config
  • MS-260 Refine log
  • MS-249 Check machine hardware during initialize
  • MS-261 Update faiss version to 1.5.3 and add BUILD_FAISS_WITH_MKL as an option
  • MS-266 Improve topk reduce time by using multi-threads
  • MS-275 Avoid sqlite logic error excetion
  • MS-278 add IndexStatsHelper
  • MS-313 add GRPC
  • MS-325 add grpc status return for C++ sdk and modify some format
  • MS-278 Add IndexStatsHelper
  • MS-312 Set openmp thread number by config
  • MS-305 Add CPU core percent metric
  • MS-310 Add milvus CPU utilization ratio and CPU/GPU temperature metrics
  • MS-324 Show error when there is not enough gpu memory to build index
  • MS-328 Check metric type on server start
  • MS-332 Set grpc and thrift server run concurrently
  • MS-352 Add hybrid index

Feature

  • MS-180 Add new mem manager
  • MS-195 Add nlist and use_blas_threshold conf
  • MS-137 Integrate knowhere

Task

  • MS-125 Create 0.3.1 release branch
  • MS-306 Optimize build efficiency

Milvus 0.3.0 (2019-06-30)

Bug

  • MS-104 Fix unittest lcov execution error
  • MS-102 Fix build script file condition error
  • MS-80 Fix server hang issue
  • MS-89 Fix compile failed, libgpufaiss.a link missing
  • MS-90 Fix arch match incorrect on ARM
  • MS-99 Fix compilation bug
  • MS-110 Avoid huge file size

Improvement

  • MS-82 Update server startup welcome message
  • MS-83 Update vecwise to Milvus
  • MS-77 Performance issue of post-search action
  • MS-22 Enhancement for MemVector size control
  • MS-92 Unify behavior of debug and release build
  • MS-98 Install all unit test to installation directory
  • MS-115 Change is_startup of metric_config switch from true to on
  • MS-122 Archive criteria config
  • MS-124 HasTable interface
  • MS-126 Add more error code
  • MS-128 Change default db path

Feature

  • MS-57 Implement index load/search pipeline
  • MS-56 Add version information when server is started
  • MS-64 Different table can have different index type
  • MS-52 Return search score
  • MS-66 Support time range query
  • MS-68 Remove rocksdb from third-party
  • MS-70 cmake: remove redundant libs in src
  • MS-71 cmake: fix faiss dependency
  • MS-72 cmake: change prometheus source to git
  • MS-73 cmake: delete civetweb
  • MS-65 Implement GetTableRowCount interface
  • MS-45 Implement DeleteTable interface
  • MS-75 cmake: change faiss version to 1.5.2; add CUDA gencode
  • MS-81 fix faiss ptx issue; change cuda gencode
  • MS-84 cmake: add arrow, jemalloc and jsoncons third party; default build option OFF
  • MS-85 add NetIO metric
  • MS-96 add new query interface for specified files
  • MS-97 Add S3 SDK for MinIO Storage
  • MS-105 Add MySQL
  • MS-130 Add prometheus_test
  • MS-144 Add nprobe config
  • MS-147 Enable IVF
  • MS-130 Add prometheus_test

Task

  • MS-74 Change README.md in cpp
  • MS-88 Add support for arm architecture

Milvus 0.2.0 (2019-05-31)

Bug

  • MS-32 Fix thrift error
  • MS-34 Fix prometheus-cpp thirdparty
  • MS-67 Fix license check bug
  • MS-76 Fix pipeline crash bug
  • MS-100 cmake: fix AWS build issue
  • MS-101 change AWS build type to Release

Improvement

  • MS-20 Clean Code Part 1

Feature

  • MS-5 Implement Auto Archive Feature
  • MS-6 Implement SDK interface part 1
  • MS-16 Implement metrics without prometheus
  • MS-21 Implement SDK interface part 2
  • MS-26 cmake. Add thirdparty packages
  • MS-31 cmake: add prometheus
  • MS-33 cmake: add -j4 to make third party packages build faster
  • MS-27 support gpu config and disable license build config in cmake
  • MS-47 Add query vps metrics
  • MS-37 Add query, cache usage, disk write speed and file data size metrics
  • MS-30 Use faiss v1.5.2
  • MS-54 cmake: Change Thrift third party URL to github.com
  • MS-69 prometheus: add all proposed metrics

Task

  • MS-1 Add CHANGELOG.md
  • MS-4 Refactor the vecwise_engine code structure
  • MS-62 Search range to all if no date specified