Releases · NVIDIA/ais-k8s

Release Highlights

Add Kubernetes Operator for deploying and maintaining AIS cluster in Kubernetes environment;
Update Helm charts to support AIS v3.4;
Add support for Terraform to spin up and manage GKE on GCP;
Provide direct external access to cluster via LoadBalancer;
Integrate Cilium CNI with Direct Server Return (DSR);
Test AIS on GKE and Minikube with and without Cilium CNI and MetalLB.

Kubernetes Operator

fix creating multiple AIS Clusters within one K8s cluster !105
port features from helm deployment !106
add integration test using envtest !107
fix hostpath create directory if not present !108
add test to check all k8s objects are created !109
add test for deploying multiple AIS clusters in the same K8s cluster !110
use a separate namespace for tests !111
tests wait for daemon statefulsets to be ready !113
update primary start-up logic; simplify init scripts !114
implement scale up logic !115
allow skipping test while building !116
single state change per reconciler loop !119
use DNS for targets and proxies !120
enable external access !121
using AIS APIs for testing !122
enabling testing on GCP + detect test environments !124
user specified storage class in tests !126
squash all operator git commits !129
set cluster config when CR updated !131
skip external tests on local minikube if tunnel not running !132
introduce event recorders !133
scale up/down with externalLB !134
manager include deployment type !135
revise state management !136
add cluster startup with external lb tests !137
implement better error management !138
cleanup RBACon deleting AIS cluster !139
fix CR cleanup !140
general code refactoring !141
tests - wait for the cluster to actually start !145
update to use new config !147
use different ConfigMap for global config !151
rewrite wait for cluster ready in tests !152
wait for resources to be truly removed on scale-down/cleanup !163
retry Reconcile on CR versions conflict !165
update code to reflect new config and cmn package !168
simplify proxy and target ready logic !169
don't update already correct state !171
sync with latest AIS master !172
add metallb makefile step !173
cleanup PVCs when reclaim policy is Delete !174
test ensuring data safety when cluster is delete !175
attempt graceful shutdown/decommission before cleaning up k8s resources !179
configurable cluster DNS domain !180
change default value of nodiskio to false !181
include liveness script !182
include readiness script !183
decommission target before scale down !184
shutdown cluster through primary !185
skip internal deployment tests !186
handle scaledown in proxies !187
setup webhooks for validation !188
implement upgrade strategy !189

Terraform

Deploy AIStore on the cloud !16
update readme + fix destroy script !17
changes in scripts and instance configuration !18
Remove VPC for ssh to work !20
Remove k8s service pinging when starting initContainer !21
minor: fix gitignore for .terraform directories !22
add admin container !23
remove nodes labels !24
update admin container !25
terraform gcp persistent storage !26
Add state for smoother redeployment and destroy of resources !28
deploy: update admin container repository and tag !29
scripts: update aisnode image and minor fixes around deployment !30
scripts: fix destroying k8s/gcp terraform volumes !31
Remove setting daemonID in uuid_env !33
improve docs !34
destroy pvc order !35
deploy: use global accessor instead of custom variable !37
add ais labels to nodes directly in gcp !38
correctly handle state var updates !39
use default node values !40
make 'machine_type' variable + minor refactor !41
use single zone instead of region to deploy GKE !42
use xfs with gcp disks !43
remove gcp AIS disks when AIS is removed !44
fix portable issue in unset_state_var !45
trap SIGINT with information to the user !47
add text about admin container !48
guard gcloud command with cloud provider check !49
add useful commands after the cluster is deployed !50
better disks removal handling !51
extend troubleshooting section in README !54
make cluster name configurable !56
add ssh key only if exists !57
fix deploy script flag parsing !58
add option to wait for the Pods to start !60
ignore error when deleting GCP disks !61
reuse persistent volumes when deleting AIS cluster !64
fix state file error message !65
terraform: persist disks on k8s cluster removal !72
deploy: check if kubectl has any nodes assigned when deploying AIStore !73
terraform: fix deleting persisted pv and pvc !74
add deploy all gif !80
fix output quotes bug !83
update aistore/aisnode image tag !86
add deploy flags for aisnode and admin containers !88

Helm

split ais chart from main ais repo !1
sync change from main repository !2
rebalance.dont_run_time missing value !3
Add missing 'allow_guest' value !4
Change ENV variables to match AIStore !5
aisloader - move chart from main repo; add automation; move playbooks !6
introduce dockerhub repo !7
initContainer must use --overwrite on pod label !8
wrap aisloader control pod in a deployment !9
Remove obsolete config option allow_guest !10
add AIS_NODE_NAME env referring to k8s node !11
aisloader to user AIS_ENDPOINT !13
Remove rproxy from config !14
Update K8s target env variable name !15
deploy: make setting AIS_HOST_IP optional !32
make graphite optional !46
correctly remove ',' at the end of the mountpaths !53
fix updating/downloading dependencies !59
reduce timeout on checking for existing cluster !78
add 'pods/log' resource for cluster role !84
parametrize admin container image !87
use target DNS !94
deploy cilium CNI !95
external access using load-balancer !96
config: ais config rename ipv4 to hostname !97
update pod management strategy statefulset !99
use target DNS instead of pod IP; update readiness probe !100
use docker images from aistore repository !101
add support for aws buckets !102
change cloud config section to backend !103
deploy proxies as StatefulSet, use DNS names as hostnames !123
role-oriented affinities, add targets anti-affinity !125
remove non-electable proxy template, enable proxies anti-affinities(https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node/) !144
update to use new config !148
add env variables for conditional primary wait !150
fix config struct !158
fix log_dir config !167

CI, docs and Misc

docs: fix supported cloud providers !36
Update README + Improve scripts !27
add terraform job running cluster on GCP !62
run CLI tests when testing terraform cluster !63
docs: update k8s and terraform readmes !66
docs: add MacOS installation guide !67
deploy: remove ETL pods on destroy !68
reduce max-host timeout !69
deploy: change ETL label !71
docs: revisit cloud deployment readme !75
Remove K8S_HOST_NAME env variable !76
docs: small improvements !77
docs: update --wait flag description and usage !79
ci: add possibility to run pipeline manually !82
docs: deploy readme cleanup !85
ci: build, use and push nightly and latest Docker images !91
ci: incorporate changes in building aistore/aisnode repo !92
enable lint and build stages for operator !117
create operator test phase !118
fix failing pipeline !130
tests: print tests logs immediately as they happen !143
remove dont_run_time config !146
add Makefile !153
organize pipeline jobs, add stub for operator GCP job !154
run operator tests on GCP !155
add skip GCP CI on skip-ci-gcp label !156
update config to reflect aistore repo !157
don't run long tests after MR is merged !160
tests: introduce short operator tests !162
show AIStore logs when running tests !166
change fspath config representation !170
Add Ansible host and config examples !176
Minor fixes, updates, and security improvements !178

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Release Highlights

Release Highlights

Kubernetes Operator

Terraform

Helm

CI, docs and Misc

Assets:

Docker Images:

Releases: NVIDIA/ais-k8s

v0.8

v0.7

v0.6

Release Highlights

AIStore Kubernetes - v0.5

Release Highlights

Kubernetes Operator

Terraform

Helm

CI, docs and Misc

Assets:

Docker Images: