v2.0.0

Changelog

All notable changes to this project will be documented in this file. Dates are displayed in UTC.

Generated by auto-changelog.

v2.0.0

2 December 2022

• do not include V1 changes in V2 changelog (#4473) #4474
• fix image build #4470
• Update docs to reference new SCv2 location #667
• Rename Go modules for repo migration #666
• Improve logging for config and log levels #453
• Remove versions dropdown and align top bar with SCv1 docs #665
• Fix nil pointer exception on scheduler restart #664
• Add ReadTheDocs config file #663
• remove unnecessary event on unload #662
• Fix http calls mirror pipelines not working #660
• Fix experiment stop #661
• Agent grpc server max connections #655
• adjust workflows for migration #654
• Fix Server replica Helm templating #648
• Add new constant rate scenario #652
• revert pipeline create changes #651
• use consistent name for consumer groups #649
• k6 docker file fixes #637
• Change to use mlserver 1.2.0 #647
• Add ability to wait on 3 termination for drain #644
• Add Helm parameterisation for server replicas #643
• Update trailer check to one if block #642
• Add huggingface capability to Helm charts #641
• only add trailers headers if not nil #640
• set parallel workers to zero for explainers and update notebooks #632
• Use kafka 3.3.1 by default #631
• Prevent terminationGracePeriodSeconds being treated as string in helm charts #628
• Run misspell -w . on docs #626
• Fix typos in pipeline docs #625
• fix type of terminationGracePeriodSeconds in helm charts #623
• Pipeline Readiness #547
• HPA server autoscale #590
• MLServer update to 1.2.0.rc5 #617
• set chart version also for seldon-core-v2-certs #620
• Add validation for empty pipeline steps #607
• Adding UnloadEnvoyRequested model replica state #616
• Add securityContext parameterisation to Kubernetes manifests for OpenShift compatibility #606
• Run all notebooks and updates to pipeline validation #597
• Model gateway logs #613
• Fix k8s version updates #612
• Respect MLServer content type in pipeline gateway http server responses #600
• Add omitempty tag to parallel_workers field #603
• Add resource parameterisation for all components in Helm chart #596
• remove duplicated seldon-v2-crds file #594
• Ensure kafka consumer reconnects happen by not ending consumers #595
• Add pipeline state to k8s resource status #591
• Fix pipeline http calls to use headers and reintroduce prom metrics removed #580
• Revert to vanilla notebook to markdown conversion #585
• Update kind and k8s versions in Ansible setup #586
• Return model name in pipeline errors #583
• Add Apache 2 Licence to code files #584
• Override config file with CLI args for Seldon CLI #579
• add rolling update md #582
• Updates for Model rolling updates #566
• Quickfix/cli pipeline error #572
• Quickfix/agent scheduler restart #568
• Issue 561 pipeline err #569
• Remove redundant pipeline gateway and agent Prometheus metrics #554
• Ensure model gateway keeps recalling scheduler and does not restart #563
• Docs fix for readthedocs static images #565
• Fix static image uris #564
• Quickfix/fixstop notready #559
• Updating styles + new images #548
• Agent startup bug fix #558
• only output warning in case of error #556
• Add missing Ansible config vars + fix incorrect docs on metric names #553
• Remove namespace from Prometheus metric names #550
• Issue 518 agent stop cmd #523
• small fixes for huggingface demo #549
• Fix initial routes for Pipelines and delete of Pipelines from Envoy #543
• cert download script and docs #544
• Add LICENSE and script to add copyright to Go files #542
• fix space in go file name #541
• Speech to Sentiment Example Updates #540
• Quickfix/add paper reference #525
• Huggingface Speech to Sentiment Example #519
• Bump MLServer version #538
• small fixes for mlserver, docs and envoy yaml #537
• [CLI] Support authority headers for control-plane subcommands #531
• Add missing Envoy patch file for Kustomize generation of Helm charts #536
• [CLI] Use positional args for resource name in CLI server-status subcommand #533
• [CLI] Use cURL-style request metadata logging for gRPC #528
• Add gRPC service name prefix in components Helm chart #530
• Add initial drift and outlier docs #529
• [CLI] Support authority headers for inference requests #526
• Use cURL-style request metadata logging in Seldon CLI #524
• Add Helm parameterisation for scheduler and Envoy service types #520
• Add a note about resetting model autoscaling #517
• Update autoscaling docs #514
• Issue 507 model autoscaling docs #513
• Add huggingface as runtime and example #511
• Install doc updates and developer doc additions #510
• Add server and certs Helm charts and raw manifests to published assets #508
• Issue 445 scheduler model autoscale #472
• Add ssl_verify_path for explainer TLS #495
• Notebook doc updates #499
• Run inference servers as non root locally #500
• Bump Kustomize to v4.5.4 #497
• Handle scheduler errors in controller and decide if retryable #484
• undeploy local before deploy local #496
• Fix Strimzi Helm values ZK indent bug + stale broker service name #492
• tidy up notebook with more models for triton #494
• revert REQUESTS_CA_BUNDLE #491
• improve CLI config load errors #489
• remove colors from outputs in batch examples #482
• Update README.md #488
• Use Helm chart for Kafka cluster setup #477
• Add batch examples to docs #481
• simplify overcommit notebook example #476
• Add log level config to dataflow engine #456
• add tritonclient example #443
• inference examples and raw contents fix #468
• Release 0.2 testing #463
• Issue 451 metrics fix #455
• fix kafka namespace: kafka -> seldon-mesh #464
• Fix missing pipeline ID in data-flow engine consumer groups #462
• Small docs updates #459
• Envoy TLS #446
• Issue 452 Fix nil deference in pipeline inspection in CLI #454
• docs update #448
• Model autoscaling (agent) #440
• Kafka SSL plus refactor of Control Plane SSL #441
• Add mTLS for data-flow engine #439
• Issue-433 Upgrade dataflow engine dependencies #438
• Remove timestamp, better verbose description #437
• Add agent mTLS #430
• Allow seldon inspect to output raw or json #432
• minor fix for k6 tests env #431
• fix hodometer docker build #428
• Fix operator Docker build #427
• fix typo #426
• fix experiment yaml #424
• Use separate Go module for generated API client #422
• Shadows #404
• Parameterized models #419
• fix experiment version bug and add notebook #420
• add missing #418
• update notebook docs #417
• Fix lazy reload #416
• fix x-seldon-route headers in pipeline chains #415
• Fix issues with versions of pipelines #414
• Update to mlserver.1.2.0-dev5 #412
• small docs update #411
• Fix experiment bug #405
• Refactor pipeline subscriber for separation of concerns #408
• Tidy Kafka config handling #407
• Scale KStream threads with pipeline steps #406
• Add control plane TLS #397
• Use Distroless image for dataflow engine #403
• Re-use topology builder for entire pipeline in dataflow engine #402
• Docker compose fixes and image size reduction #401
• fix bugs with container dockerfiles #398
• Issue-393 Use Distroless images for Go apps #396
• Fix helm docs and updating linting #392
• Namespaced controller #380
• Revert "k8s codegen script and generated client (#377)" #391
• update cifar10 example #390
• Fix scheduler PVC volume not writable on GKE #387
• Issue-373 Remove irrelevant k8s metrics from Hodometer #381
• k8s codegen script and generated client #377
• Add per pipeline histogram metrics #382
• Fix: update k8s CRDs after experiment change #379
• income example #378
• Passing parameters to batched requests from data-flow engine #374
• Enable pipelinegateway multi topics consumer #372
• Refactor metrics and add separate pipeline and model metrics #371
• Enable kafka kraft in k8s #370
• Modelgateway topic to consumer consistent hashing #368
• Pipeline Experiments #360
• Read custom server example #361
• Change admin client create for create topics #359
• Various fixes for modelgateway usage at scale #358
• Add mlflow model in k6 examples #349
• fix sticky session usage with header addtion in lua #354
• Add Makefile support for Compose build param #356
• Add Hodometer service dependency on scheduler in Compose #355
• add envs for otel enable in docker compose #347
• Use right image for hodometer and add helper for pulling images #350
• Add usage metrics (Hodometer) docs #343
• Explainers #298
• add release process description #329
• fix trigger joins #340
• add hodometer to the list of images to be built by GA #344
• Move CHANGELOG.md to top level #342
• Add hodometer deployments #318
• v0.1.0 change log #337
• Fix pipelinegateway panic upon kafka reconnect #339
• Add sticky sessions for experiments #250
• Fix shm config for loading python model on triton (k8s) #338
• Ignore RC builds in generated core-release notes #336
• Update README.md #330
• V2 release process #327
• add missing kustomize patch + version setting helpers #322
• Remove duplicated Compose image tags in Makefile #325
• add img overrides for compose #324
• Ensure X-Request-ID is returned and allow CLI inspect to use #314
• Add container merge sematics for easier custom servers #315
• Add server status update batching #307
• Sherif akoush/demo fixes #312
• Add build info to Hodometer Docker images #304
• Add batching for XDS server updates #248
• Add Kubernetes metrics to Hodometer #299
• Modelgateway issue #296
• Experiment store #292
• Sherif akoush/upgrade mlserver 1.1 #294
• Sherif akoush/http lazy load fix #293
• Sherif akoush/lazy load model in restart #291
• Num modelgateway workers fix #290
• Robustness fixes #287
• Modelgateway workers from envar #288
• Sherif akoush/create snapshot optimisation #285
• Modelgateway threads #278
• Fix trigger join #282
• Pipeline db fixes #283
• Allow model scaling, k6 constant throughput tests and Prometheus/Grafana in Docker Compose install #262
• Sherif akoush/improve replica sorting #280
• do not run action to build/push images on forks #269
• fix workflow dispatch inputs for image building #268
• Change Envoy LB Algorithm to Least Requests #265
• Request metadata #264
• Use static Kafka consumer in Kstreams #260
• Scheduler db folder docker #261
• Tracing config #225
• Separate event publish from locked updates to data structures #254
• add mnt folder to git with .keep file #255
• small update to docs #253
• Dataflow doc #245
• Update index.md #249
• update cifar10 demo #247
• Metrics dashboard docs #244
• fix http reverse proxy port issue #243
• add workflow that builds and push images #234
• Cli command updates #242
• Sherif akoush/metrics dashboard #241
• add longer default timeout in Envoy configuration #239
• Optimise XDS server route creation #237
• v2 control plane grpc + various fixes for scalability #229
• Some text updates, new reference #236
• Update index.md #235
• Pipeline Persistence #188
• Add locks to streams #231
• small docs title page update #228
• add locks around stream send #226
• docs updates #227
• Add Hodometer stub receiver #211
• update cli export flags docs #224
• Experiment status #212
• fix trigger NullPointer exception #221
• docs update #222
• fix external port for kafka when running internally #220
• Change port to avoid conflict #219
• read events in go routines #215
• Update server snapshot creation in scheduler #214
• Sherif akoush/fix v2client load #210
• Sherif akoush/add evict metrics #184
• CIFAR10 Example and CLI Pipeline inspect #207
• Add short names to CRDs #209
• update pipeline status and finalizer check #208
• Cli updates #199
• Helm chart creation #193
• Kafka configuration via config files #189
• Alibi-detect iris drift detection example #191
• throw errors in CLI on bad yaml #198
• Add usage metrics collector #181
• Add model metadata to CLI #187
• update install docs for Ansible #190
• Allow max message size in kafka and grpc #186
• change kafka to bitnami #185
• Allow for more informative scheduling errors #182
• Add kafka Produce and Consume Tracing Spans #178
• Sherif akoush/report same message from agent #183
• Requests batch processing transformer implementation #135
• fix docker compose for host network #177
• Ansible: add jaeger and opentelemetry #172
• Add server extra capabilities #169
• Update memory.go #173
• Parametrize ansible #167
• Update Configuration.kt #166
• add inference docs #165
• initial ansible playbooks #154
• Further Docs #163
• Open Telemetry Tracing #160
• make pipeline different from modelname #162
• Sherif akoush/k6 dataflow #149
• Pipeline inputs #158
• fix locks in pipeline state set #159
• Add Pipeline Triggers #152
• tutorial docs section #156
• Allow local model folders #153
• Docs update #150
• Use Gradle directly in dataflow engine Docker build #148
• allow both mlserver and triton to be started locally #146
• Add clearer state logging for chainer and joiner #145
• Add state listeners to kstreams to wait while rebalancing takes place #144
• Conditional and Error Pipelines #143
• Pipelines on k8s #142
• Added install command for local examples notebook #141
• Dataflow updates #140
• fix dataflow bugs #139
• Update dataflow joiner #137
• Update golangci-lint to 1.45.2 #138
• Pipelines with Join #136
• Add features page placeholder #133
• Add Docker setup for data-flow engine #134
• Docs - add k8s resources #132
• Docs Draft Outline #131
• Fix data-flow Gradle setup #130
• Add Kafka Streams data-flow engine #119
• pick free port in test #128
• add mlserver protos extensions back #127
• Persisting k6 results to GCS bucket #125
• Add an outline of software design doc #126
• rename stream to modelgateway #124
• Sherif akoush/various fixes for testing #117
• initial docs setup #123
• CLI plus updated sample notebooks #120
• Disable auto-loading of models in MLServer at start-up via env vars #122
• Pipeline operator #118
• Pipelines #107
• Wire up overcommit with scheduler #111
• proto update for chainer #109
• Experiments v1 #106
• chainer protos #108
• Sherif akoush/simplify locks #103
• Sherif akoush/remove version code agent #102
• Stream integration with Kafka #104
• Add event bus for scheduler-internal events #99
• Fix maybe parsing methods so not a fatal on not found #96
• Refactor Agent cmd package and argument parsing #83
• replace loaded models with versioned models key #88
• Add memory sorter to default scheduler #91
• fix rclone host docker compose config #94
• Sherif akoush/update v2 protos in notebooks #93
• Sherif akoush/scv2 50/flatten versions (and various other fixes) #86
• Prometheus Inference Metrics #82
• Traffic split envoy #79
• Docker compose updates #84
• Use Compose for Docker-based Make targets #81
• Wiring up proxies #78
• Add Docker Compose manifests #77
• Sherif akoush/reverse proxy grpc 2 #72
• K6 Load Tests #69
• Service Mesh experiments Istio, Traefik, Ambassador #70
• Sherif akoush/Memory over-commit (reverse proxy) #18
• Server Custom Resource #41
• Versioning #40
• Move Protobuf contracts to top-level #39
• Initial Operator update for Model resource #20
• Add scheduler proxy/stub #38
• Dynamic RClone Configuration #8
• Format the code for new lines at end of file (gofmt) #11
• Add github actions for linting and tests #10
• Add golangci-lint linters and fix existing lint failures #9
• Add payload logging with Envoy Taps #1
• Updated to non deprecated grpc settings #29
• add gRPC inference #28
• Seldon Core V2 Scheduler Update #27
• Add V2 APIs and Samples #26
• New Operator APIs #25
• Remove initial operator #24
• SCV2 POC Update #23
• Seldon Core V2 - Scheduler experiments #22
• Seldon core v2 (add smoke test) #21
• Seldon core v2 (further updates) #20
• Generating changelog for v2.0.0 014a935
• Seldon V2 APIs initial Draft 02d963c
• Generating changelog for v2.0.0 d955a61
• Initial commit for Model reconcile a1dfb6d
• update status for model d127e19