fix(image): merges manifests builder stages to one #773
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
bartoszmajsak
requested review from
zdtsw and
ykaliuta
and removed request for
etirelli and
LaVLaS
We can combine two build stages to one, as there is no need to always build both images to only then decide from which one we want to copy manifests to the target image. Instead `manifests` stage will either copy local manifests or fetches using the script based on `FETCH_MANIFESTS` argument (previously known as `USE_LOCAL`). In addition it opens up `IMAGE_BUILD_FLAGS` variable to be overwritten when executing make targets instead of changing its value in the `Makefile` directly, as it was previously needed.
bartoszmajsak
force-pushed
the
builder_img_stages
branch
from
db658ef to
b7735c3
Compare
ykaliuta
reviewed
Dec 1, 2023
Makefile
Outdated
| # set to "true" to use local instead | ||
| # see target "image-build" | ||
| IMAGE_BUILD_FLAGS = --build-arg USE_LOCAL=false | ||
| IMAGE_BUILD_FLAGS ?= --build-arg FETCH_MANIFESTS=true |
There was a problem hiding this comment.
It does not make sense for the usage like in the README.md. Command line variables have priority in make anyway. Even over environment regardless of -e flag.
But may be having a separate variable to avoid passing --build-arg can be a good idea, what do you think?
There was a problem hiding this comment.
It does not make sense for the usage like in the README.md. Command line variables have priority in make anyway. Even over environment regardless of -e flag.
That's right. Thanks for catching it. I think the source of confusion for me was not expecting to see both build stages (git clone and COPY) executed in the original form of the docker file when I was expecting to see only the one based on the flag. That's now cleaned up in 0d3a574
But may be having a separate variable to avoid passing --build-arg can be a good idea, what do you think?
I am on the verge here. I have
alias image='make image -e IMAGE_BUILD_FLAGS="--build-arg FETCH_MANIFESTS=false" -e IMG=quay.io/maistra-dev/opendatahub-operator:$\(VERSION\) -e IMAGE_BUILDER=docker -e VERSION=dev-$(git branch --show-current)'so I'm good.
Would it make it easier for you to have such a flag?
There was a problem hiding this comment.
But may be having a separate variable to avoid passing --build-arg can be a good idea, what do you think?
I am on the verge here. I have
alias image='make image -e IMAGE_BUILD_FLAGS="--build-arg FETCH_MANIFESTS=false" -e IMG=quay.io/maistra-dev/opendatahub-operator:$\(VERSION\) -e IMAGE_BUILDER=docker -e VERSION=dev-$(git branch --show-current)'so I'm good.
Great! Just keep in mind that -e is redundant here, it works differently for make than to podman-run.
Would it make it easier for you to have such a flag?
No actually, I haven't ever used that in practice :). So I personally would vote to avoid the change at all. Especially because it's unrelated to the purpose of the patch.
There was a problem hiding this comment.
@bartoszmajsak was thinking about avoiding go version setting in 4 places. One of the ideas was getting it from go.mod and instead of patching Dockerfile, pass it as the argument (similar to toolbox), but then your alias will be broken.
Dockerfiles/Dockerfile
Outdated
| ARG GOLANG_VERSION=1.19 | ||
| ARG USE_LOCAL=false | ||
| ARG FETCH_MANIFESTS=true |
There was a problem hiding this comment.
Does it make sense to leave it here (except as a matter of documentation) since according to the Dockerfile documentation ARG is valid till the next FROM (USE_LOCAL was expanded in the FROM line itself before)?
There was a problem hiding this comment.
That's right, I cleared it up in 51d8c5d
|
It actually did not build both images (if I understand my build logs correctly) due to dependency tracking. But the change makes Dockerfile much cleaner. A question I have -- does it make sense to avoid that manifest stage and do it in the bulider? It used go-toolset image to make manifests before. |
@ykaliuta Ok I'm triple puzzled now, have a look at the build log below docker builddocker versionClient: Version: 20.10.17 API version: 1.41 Go version: go1.16.15 Git commit: aa7e414 Built: Fri Jul 8 23:32:01 2022 OS/Arch: linux/amd64 Context: default Experimental: true |
Hah, interesting. Looks like podman is more clever. But I remember I read about dependencies in docker as well in the docs :) Makes more sense then. But what about squashing it one phase then? AFAIU it was splitted into two phases to make that condition on the image builder level. With the patch when it's in the script there is no sense for additional phase. Or I'm missing something? podman log```podman build --no-cache -f Dockerfiles/Dockerfile --build-arg USE_LOCAL=false -t quay.io/ykaliuta/opendatahub-operator:latest . [1/4] STEP 1/6: FROM registry.access.redhat.com/ubi8/go-toolset:1.19 AS builder_local_false [1/4] STEP 2/6: ARG OVERWRITE_MANIFESTS --> 0658cce527fa [1/4] STEP 3/6: USER root --> 6d2a0b9363d0 [1/4] STEP 4/6: WORKDIR /opt --> 3bd13acfa0fb [1/4] STEP 5/6: COPY get_all_manifests.sh get_all_manifests.sh --> fc074ef2b318 [1/4] STEP 6/6: RUN ./get_all_manifests.sh ${OVERWRITE_MANIFESTS} Cloning repo kserve: opendatahub-io:kserve:release-v0.11.0:config:kserve Cloning into './.kserve'... Cloning repo odh-dashboard: opendatahub-io:odh-dashboard:incubation:manifests:dashboard Cloning into './.odh-dashboard'... Cloning repo odh-notebook-controller: opendatahub-io:kubeflow:v1.7-branch:components/odh-notebook-controller/config:odh-notebook-controller/odh-notebook-controller Cloning into './.kubeflow'... Cloning repo notebooks: opendatahub-io:notebooks:main:manifests:notebooks Cloning into './.notebooks'... Cloning repo odh-model-controller: opendatahub-io:odh-model-controller:release-0.11.0:config:odh-model-controller Cloning into './.odh-model-controller'... Cloning repo ray: opendatahub-io:kuberay:master:ray-operator/config:ray Cloning into './.kuberay'... Cloning repo kf-notebook-controller: opendatahub-io:kubeflow:v1.7-branch:components/notebook-controller/config:odh-notebook-controller/kf-notebook-controller Cloning into './.kubeflow'... Cloning repo data-science-pipelines-operator: opendatahub-io:data-science-pipelines-operator:main:config:data-science-pipelines-operator Cloning into './.data-science-pipelines-operator'... Cloning repo trustyai: trustyai-explainability:trustyai-service-operator:release/1.10.2:config:trustyai-service-operator Cloning into './.trustyai-service-operator'... Cloning repo model-mesh: opendatahub-io:modelmesh-serving:release-0.11.0:config:model-mesh Cloning into './.modelmesh-serving'... Cloning repo codeflare: opendatahub-io:codeflare-operator:main:config:codeflare Cloning into './.codeflare-operator'... --> a0196c56ddc5 [3/4] STEP 1/13: FROM a0196c56ddc53a2f62783f1a40a0ecc47ec0c05055fb7f7dbe1afa161d9ede1d AS builder [3/4] STEP 2/13: USER root --> 92530b5fda6a [3/4] STEP 3/13: WORKDIR /workspace --> 262e23812a42 [3/4] STEP 4/13: COPY go.mod go.mod --> e39a33962a8b [3/4] STEP 5/13: COPY go.sum go.sum --> d145cda8323c [3/4] STEP 6/13: RUN go mod download --> 4a2bee6f7ac5 [3/4] STEP 7/13: COPY apis/ apis/ --> 432be0644019 [3/4] STEP 8/13: COPY components/ components/ --> 8d409afb27fd [3/4] STEP 9/13: COPY controllers/ controllers/ --> b046471d1ddf [3/4] STEP 10/13: COPY main.go main.go --> 35582a002821 [3/4] STEP 11/13: COPY pkg/ pkg/ --> 2c2841d130be [3/4] STEP 12/13: COPY infrastructure/ infrastructure/ --> 405d2af8cba9 [3/4] STEP 13/13: RUN CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -a -o manager main.go --> b4b01e4fb965 [4/4] STEP 1/7: FROM registry.access.redhat.com/ubi8/ubi-minimal:latest [4/4] STEP 2/7: WORKDIR / --> 5a1ddd47a55a [4/4] STEP 3/7: COPY --from=builder /workspace/manager . --> 4873e93cdc67 [4/4] STEP 4/7: COPY --chown=1001:0 --from=builder /opt/odh-manifests /opt/manifests --> 4fc738ad01f0 [4/4] STEP 5/7: RUN chown -R 1001:0 /opt/manifests && chmod -R a+r /opt/manifests --> 984a23fbba03 [4/4] STEP 6/7: USER 1001 --> ff88d059ae68 [4/4] STEP 7/7: ENTRYPOINT ["/manager"] [4/4] COMMIT quay.io/ykaliuta/opendatahub-operator:latest --> e154b0975db4 Successfully tagged quay.io/ykaliuta/opendatahub-operator:latest e154b0975db48709bbcb7aa5b5f3f8da65fddbf404beb7bde816a6c6b08e2328 ``` |
I assumed podman is clever, but I'm on a very outdated fedora on my main workstation and podman ain't happy about it :)
Let's not start this squash discussion again! Just kidding :D Personally, I like such separation, but if we significantly trim on execution time and image sizes, then I'm happy to do it. WDYT @zdtsw? |
|
/test opendatahub-operator-e2e |
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: zdtsw The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
/lgtm |
|
/test opendatahub-operator-e2e |
|
New changes are detected. LGTM label has been removed. |
|
/retest |
|
/test opendatahub-operator-e2e |
|
@bartoszmajsak: The following tests failed, say
Full PR test history. Your PR dashboard. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here. |
|
The PR is still in my github list of assignment. It was approved, looks good to me as well, but was not merged due to failing tests. |
|
@ykaliuta Unfortunately, I don't have the bandwidth to dive into a year-old PR at the very moment. That said, I’d like to know if this approach is still worth it. If it does I can squeeze in some time in the coming days. |
ykaliuta
added a commit
to ykaliuta/opendatahub-operator
that referenced
this pull request
Nov 19, 2024
We can combine two build stages into one, as there is no need to always build both images (not done by podman) to only then decide from which one we want to copy manifests to the target image. Instead manifests stage will either copy local manifests or fetches using the script based on USE_LOCAL argument. Move USE_LOCAL and OVERWIRTE_MANIFESTS args under FROM since args have scope of the FROM they are declared in. It requires opt/manifests directory to exist, but since it's a part of git repo, it's fine. Original patch from: Bartosz Majsak <[email protected]> [1] [1] opendatahub-io#773 Signed-off-by: Yauheni Kaliuta <[email protected]>
5 tasks
Well, you are the customer of that, with podman nobody noticed :) Ok, I fetched the main change into #1381 avoiding interface changes. |
auto-merge was automatically disabled
November 19, 2024 17:09
Pull request was closed
ykaliuta
added a commit
to ykaliuta/opendatahub-operator
that referenced
this pull request
Nov 19, 2024
We can combine two build stages into one, as there is no need to always build both images (not done by podman) to only then decide from which one we want to copy manifests to the target image. Instead manifests stage will either copy local manifests or fetches using the script based on USE_LOCAL argument. Move USE_LOCAL and OVERWIRTE_MANIFESTS args under FROM since args have scope of the FROM they are declared in. It requires opt/manifests directory to exist, but since it's a part of git repo, it's fine. Original patch from: Bartosz Majsak <[email protected]> [1] [1] opendatahub-io#773 Signed-off-by: Yauheni Kaliuta <[email protected]>
openshift-merge-bot bot
pushed a commit
that referenced
this pull request
Nov 20, 2024
We can combine two build stages into one, as there is no need to always build both images (not done by podman) to only then decide from which one we want to copy manifests to the target image. Instead manifests stage will either copy local manifests or fetches using the script based on USE_LOCAL argument. Move USE_LOCAL and OVERWIRTE_MANIFESTS args under FROM since args have scope of the FROM they are declared in. It requires opt/manifests directory to exist, but since it's a part of git repo, it's fine. Original patch from: Bartosz Majsak <[email protected]> [1] [1] #773 Signed-off-by: Yauheni Kaliuta <[email protected]>
zdtsw
pushed a commit
to zdtsw-forking/opendatahub-operator
that referenced
this pull request
Nov 28, 2024
We can combine two build stages into one, as there is no need to always build both images (not done by podman) to only then decide from which one we want to copy manifests to the target image. Instead manifests stage will either copy local manifests or fetches using the script based on USE_LOCAL argument. Move USE_LOCAL and OVERWIRTE_MANIFESTS args under FROM since args have scope of the FROM they are declared in. It requires opt/manifests directory to exist, but since it's a part of git repo, it's fine. Original patch from: Bartosz Majsak <[email protected]> [1] [1] opendatahub-io#773 Signed-off-by: Yauheni Kaliuta <[email protected]> (cherry picked from commit c1671ab)
zdtsw
pushed a commit
to zdtsw-forking/opendatahub-operator
that referenced
this pull request
Nov 29, 2024
We can combine two build stages into one, as there is no need to always build both images (not done by podman) to only then decide from which one we want to copy manifests to the target image. Instead manifests stage will either copy local manifests or fetches using the script based on USE_LOCAL argument. Move USE_LOCAL and OVERWIRTE_MANIFESTS args under FROM since args have scope of the FROM they are declared in. It requires opt/manifests directory to exist, but since it's a part of git repo, it's fine. Original patch from: Bartosz Majsak <[email protected]> [1] [1] opendatahub-io#773 Signed-off-by: Yauheni Kaliuta <[email protected]> (cherry picked from commit c1671ab)
zdtsw
pushed a commit
to zdtsw-forking/opendatahub-operator
that referenced
this pull request
Dec 2, 2024
We can combine two build stages into one, as there is no need to always build both images (not done by podman) to only then decide from which one we want to copy manifests to the target image. Instead manifests stage will either copy local manifests or fetches using the script based on USE_LOCAL argument. Move USE_LOCAL and OVERWIRTE_MANIFESTS args under FROM since args have scope of the FROM they are declared in. It requires opt/manifests directory to exist, but since it's a part of git repo, it's fine. Original patch from: Bartosz Majsak <[email protected]> [1] [1] opendatahub-io#773 Signed-off-by: Yauheni Kaliuta <[email protected]> (cherry picked from commit c1671ab)
openshift-merge-bot bot
pushed a commit
that referenced
this pull request
Dec 2, 2024
* update: owners (#1376) Signed-off-by: Wen Zhou <[email protected]> (cherry picked from commit 4a15a8f) Signed-off-by: Wen Zhou <[email protected]> * Dockerfile, Makefile: make CGO_ENABLED configurable (#1382) The commit a107703 ("feat(fips): enable GO_ENABLED in build (#1001)") enabled CGO which makes problems for builders on non-x86 platforms. Make it as an in the Dockerfile keeping default the same (enabled), but make it possible to override with either environment (`export CGO_ENABLED=0`) or make (`make CGO_ENABLED=0 image-build`) Signed-off-by: Yauheni Kaliuta <[email protected]> (cherry picked from commit 2d94349) * Makefile: make USE_LOCAL overridable (#1384) Get USE_LOCAL image build flag from makefile variable to make it overridable with `make USE_LOCAL=true image` Do not allow to get its value for environment be default due to pretty generic name. This is shorter than the old recommendation of overriding IMAGE_BUILD_FLAGS. And since now CGO_ENABLED is also a flag, does not mess up with it. * USE_LOCAL=true uses existing manifests from opt/manifests for the produced image without downloading them with get_all_manifests.sh making it possible to both save time and make local amendments. Signed-off-by: Yauheni Kaliuta <[email protected]> (cherry picked from commit cea41dc) * Dockerfile: merges manifests builder stages to one (#1381) We can combine two build stages into one, as there is no need to always build both images (not done by podman) to only then decide from which one we want to copy manifests to the target image. Instead manifests stage will either copy local manifests or fetches using the script based on USE_LOCAL argument. Move USE_LOCAL and OVERWIRTE_MANIFESTS args under FROM since args have scope of the FROM they are declared in. It requires opt/manifests directory to exist, but since it's a part of git repo, it's fine. Original patch from: Bartosz Majsak <[email protected]> [1] [1] #773 Signed-off-by: Yauheni Kaliuta <[email protected]> (cherry picked from commit c1671ab) * fix: rhoai build has extra manifests than odh's - changes made in odh, missing manifests for rhoai since it is only using builder for main and others from manifests stage Signed-off-by: Wen Zhou <[email protected]> --------- Signed-off-by: Wen Zhou <[email protected]> Co-authored-by: Yauheni Kaliuta <[email protected]>
zdtsw
added a commit
that referenced
this pull request
Dec 6, 2024
* cluster_config: init cluster variables on startup (#1059) * cluster_config: move type definitions to the beginning of the file Just a bit more clarity Signed-off-by: Yauheni Kaliuta <[email protected]> * cluster_config: init cluster variables on startup Cluster configuration is supposed to be the same during operator pod lifetime. There is no point to detect it on every reconcile cycle keeping in mind that it can causes many api requests. Do the initialization on startup. Keep GetOperatorNamespace() returning error since it defines some logic in the DSCInitialization reconciler. Automatically called init() won't work here due to need to check error of the initialization. Wrap logger into context and use it in the Init() like controller-runtime does [1][2]. Save root context without the logger for mgr.Start since "setup" logger does not fit normal manager's work. Leave GetDomain() as is since it uses OpenshiftIngress configuration which is created when DSCInitialization instantiated. Log cluster configuration on success from Init, so remove platform logging from main. [1] https://github.com/kubernetes-sigs/controller-runtime/blob/38546806f2faf5973e3321a7bd5bb3afdbb5767d/pkg/internal/controller/controller.go#L297 [2] https://github.com/kubernetes-sigs/controller-runtime/blob/38546806f2faf5973e3321a7bd5bb3afdbb5767d/pkg/internal/controller/controller.go#L111 Signed-off-by: Yauheni Kaliuta <[email protected]> * cluster: do not return error from GetRelease GetRelease return values are defined at startup, the error checked in main, so no point to return error anymore. Signed-off-by: Yauheni Kaliuta <[email protected]> --------- Signed-off-by: Yauheni Kaliuta <[email protected]> * components: move params.env image updating to Init stage (#1191) * components, main: add Component Init method Add Init() method to the Component interface and call it from main on startup. Will be used to move startup-time code from ReconcileComponent (like adjusting params.env). Signed-off-by: Yauheni Kaliuta <[email protected]> * components: move params.env image updating to Init stage Jira: https://issues.redhat.com/browse/RHOAIENG-11592 Image names in environment are not supposed to be changed during runtime of the operator, so it makes sense to update them only on startup. If manifests are overriden by DevFlags, the DevFlags' version will be used. The change is straight forward for most of the components where only images are updated and params.env is located in the kustomize root directory, but some components (dashboard, ray, codeflare, modelregistry) also update some extra parameters. For them image part only is moved to Init since other updates require runtime DSCI information. The patch also changes logic for ray, codeflare, and modelregistry in this regard to update non-image parameters regardless of DevFlags like it was changed in dashboard recently. The DevFlags functionality brings some concerns: - For most components the code is written such a way that as soon as DevFlags supplied the global path variables are changed and never reverted back to the defaults. For some (dashboard, trustyai) there is (still global) OverridePath/entryPath pair and manifests reverted to the default, BUT there is no logic of transition. - codeflare: when manifests are overridden namespace param is updated in the hardcoded (stock) path; This logic is preserved. Signed-off-by: Yauheni Kaliuta <[email protected]> --------- Signed-off-by: Yauheni Kaliuta <[email protected]> * update: remove webhook service from bundle (#1275) - we do not need it in bundle, CSV auto generate one during installation - if we install operator via OLM, webhook service still get used from config/webhook/service.yaml Signed-off-by: Wen Zhou <[email protected]> * update: add validation on application and monitoring namespace in DSCI (#1263) Signed-off-by: Wen Zhou <[email protected]> * logger: add zap command line switches (#1280) Allow to tune preconfigured by --log-mode zap backend with standard zap command line switches from controller-runtime (zap-devel, zap-encoder, zap-log-level, zap-stacktrace-level, zap-time-encoding)[1]. This brings flexibility in logger setup for development environments first of all. The patch does not change default logger setup and does not change DSCI override functionality. [1] https://sdk.operatorframework.io/docs/building-operators/golang/references/logging Signed-off-by: Yauheni Kaliuta <[email protected]> * Modify Unit Tests GitHub Actions workflow to get code coverage test reports (#1279) * Create codecov.yml * Added to run test coverage also on PRs * Removing trailing ] * Update .github/workflows/codecov.yml Co-authored-by: Wen Zhou <[email protected]> * Removed go mod install dependency * Consolidated codecov workflow into unit tests workflow --------- Co-authored-by: Wen Zhou <[email protected]> * webhook: move initialization inside of the module (#1284) Add webhook.Init() function and hide webhook setup inside of the module. It will make it possible to replace Init with a NOP (no operation) with conditional compilation for no-webhook build. Signed-off-by: Yauheni Kaliuta <[email protected]> * feat: pass platform from env variable or fall back to use old logic (#1252) * feat: pass platform from env variables or fall back to use old logic - introduce new env var ODH_PLATFORM_TYPE, value set during build time - if value not match, fall back to detect managed then self-managed - introduce new env var OPERATOR_RELEASE_VERSION, value also set during build time - if value is empty, fall back to use old way from CSV to read version or use 0.0.0 - add support from makefile - use envstubst to replace version Signed-off-by: Wen Zhou <[email protected]> * update: remove release version in the change Signed-off-by: Wen Zhou <[email protected]> --------- Signed-off-by: Wen Zhou <[email protected]> * fix: update release version in DSCI and DSC .status for upgrade case (#1287) - DSCI: if current version is not matching, update it - DSC: in both reconcile pass and fail case, update it Signed-off-by: Wen Zhou <[email protected]> * Update version to 2.19.0 (#1291) Co-authored-by: VaishnaviHire <[email protected]> * Makefile, webhook: add run-nowebhook (#1286) Make it possible to compile operator without webhook enabled (with -tags nowebhook). Create a stub webhook.Init() function for that. Add run-nowebhook target to run webhook locally. It requires `make install` to be executed on the cluster beforehand. Since it repeats `make run`, move the command to a variable. Also use variable RUN_ARGS for operator arguments. It makes it possible to override them from the command line. In VSCode it is possible to debug it with the following example launch.json configuration: ``` { "name": "Debug Operator No Webhook", "type": "go", "request": "launch", "mode": "debug", "program": "main.go", "buildFlags": [ "-tags", "nowebhook" ], "env": { "OPERATOR_NAMESPACE": "opendatahub-operator-system", "DEFAULT_MANIFESTS_PATH": "./opt/manifests" }, "args": [ "--log-mode=devel" ], "cwd": "${workspaceFolder}", } ``` Signed-off-by: Yauheni Kaliuta <[email protected]> * update: use namespace dyniamically from operator env than hardcode value (#1298) - thses are only needed when it is downstream speicific cases Signed-off-by: Wen Zhou <[email protected]> * webhook: switch to contextual logger (#1295) Use k8s contextual logger instead of own. Comparing to other parts of the operator webhook is a bit special since it serves own endpoints and has context independent from controllers. Include more info (operation), just to get more details. Do not add kind since it's clear from "resource" field. Since controller-framework adds "admission" to the name, use own LogConstructor with own base logger for the framework's DefaultLogConstructor. Add name locally to distinguish from framework's messages. Add Name field to the structures to avoid copying string literal for both WithName() and WithValues(). The output changes and it looks like ``` {"level":"info","ts":"2024-10-11T05:17:20Z","logger":"ValidatingWebhook","msg":"Validation request","object":{"name":"default-dsci"},"namespace":"","name":"default-dsci","resource":{"group":"dscinitialization.opendatahub.io","version":"v1","resource":"dscinitializations"},"user":"kube:admin","requestID":"e5bf3768-6faa-4e14-9004-e54ee84ad8b7","webhook":"ValidatingWebhook","operation":"CREATE"} ``` or for the defaulter: ``` {"level":"info","ts":"2024-10-11T04:50:48Z","logger":"DefaultingWebhook","msg":"Defaulting DSC","object":{"name":"default-dsc"},"namespace":"","name":"default-dsc","resource":{"group":"datasciencecluster.opendatahub.io","version":"v1","resource":"datascienceclusters"},"user":"kube:admin","requestID":"c9213ff3-80ee-40c0-9f15-12188dece68e","webhook":"DefaultingWebhook"} ``` (the messages are not from the current codebase, was added for demo only) Signed-off-by: Yauheni Kaliuta <[email protected]> * chore: use client.OjectKeyFromObject in client.Get() (#1301) Signed-off-by: Wen Zhou <[email protected]> * (fix): Change on trigger event to `pull_request_target` in the "Check config and readme updates / Ensure generated files are included (pull_request)" action to fix "Resource not accessible by integration" error while running the action (#1296) Signed-off-by: AJAY JAGANATHAN <[email protected]> * feat: Operator disable create usergroup if detect user enabled external auth (#1297) * feat: Operator disable create usergroup if detect users enabled external auth - use internal Authentication CR Type "" or IntegratedOAuth to indicate if Operator should create usergroup or not CRD has validation to only allow "IntegratedOAuth", "", "None" or "OIDC" - only grant "get, watch , list" as least permission - remove duplicated rbac for "ingress", has been defined in other places - add object into cache - add CRD into unit-test - add unit-test: since we dont have auth CR, it should not create usergroup Signed-off-by: Wen Zhou <[email protected]> * update: review comments Signed-off-by: Wen Zhou <[email protected]> * update: review comments use const Signed-off-by: Wen Zhou <[email protected]> --------- Signed-off-by: Wen Zhou <[email protected]> * components, logger: use contextual logging approach (#1253) Switch ReconcileComponent from passing logger explicitly to wrapping it into context[1][2] Makes one parameter less to pass and will allow called utilities to report component context where they are called from. No user or logging format impact until utilities takes contextual logging in use. [1] https://www.kubernetes.dev/blog/2022/05/25/contextual-logging/ [2] https://github.com/kubernetes/enhancements/blob/master/keps/sig-instrumentation/3077-contextual-logging/README.md Credits-To: Bartosz Majsak [email protected] Signed-off-by: Yauheni Kaliuta <[email protected]> * logger, controllers: use common logging level, INFO by default (#1289) Remove increasing logging level for controllers (it was also passed to components if not overridden from DSCI) since: - it made logging inconsistent. The base contoller runtime logger is set up with INFO level for all log modes, so when controllers are configured for Error level, user sees INFO messages from both controller-runtime and other parts of operator which use controller-runtime's logger directly; - since the base logger is configured for INFO, there is no difference in levels between "default" and "devel". Having levels 1 and 2 there is misleading. Update documentation. This patch changes default logging, former filtered Info messages are displayed now. There is no _big_ difference in practice since currently the log is anyway full of info messages from parts which do not use reconciler's logger, like: {"level":"info","ts":"2024-10-09T13:23:11Z","msg":"waiting for 1 deployment to be ready for dashboard"} Signed-off-by: Yauheni Kaliuta <[email protected]> * removed lint and typo fixes (#1302) * fix: to make the unset env variable in CSV work with fallback (#1306) - previous code, will run into opendathaub case if env is not set Signed-off-by: Wen Zhou <[email protected]> * controllers: switch to k8s contextual logger (#1308) Jira: https://issues.redhat.com/browse/RHOAIENG-14096 This is a squashed commit of the following patches: d012b67 ("controllers: switch to k8s contextual logger") 9880530 ("logger: blindly convert ctrl.Log users to contextual") - controllers: switch to k8s contextual logger Remove own logger from controllers' reconcilers and switch to k8s contextual logger instead [1]. Use contextual logger for SecretGeneratorReconciler and CertConfigmapGeneratorReconciler setups as well. Add name to the logger coming from the framework. It will contains "controller" field already, and like in webhook with the name it's easy to distinguish framework and operator messages. - logger: blindly convert ctrl.Log users to contextual All the users should have proper context now. The log level changes will affect it as well. [1] https://www.kubernetes.dev/blog/2022/05/25/contextual-logging/ Signed-off-by: Yauheni Kaliuta <[email protected]> * Tidy up run-nowebhook recipe & make clean PHONY (#1310) * chore: Tidy run-nowebhook recipe The suggestion to tidy up the run-nowebhook recipe comes from this conversation: https://github.com/opendatahub-io/opendatahub-operator/pull/1304/files#r1806731373 * chore: Make `clean` a PHONY target I believe `clean` should be a PHONY target, since it doesn't create a file called `clean` * refactor the secret generator controller (#1311) The commit is meant to: - make the code easier to understand, reducing complexity caused by nested if/else and error conditions (align happy path to the left) - remove shadowed error vars to avoid reporting misleading errors - add some basic unit test for the reconcile loop * update: rename variables rhods to rhoai (#1313) Signed-off-by: Wen Zhou <[email protected]> Co-authored-by: ajaypratap003 <[email protected]> * logger: add logLevel to DSCI devFlags (#1307) Jira: https://issues.redhat.com/browse/RHOAIENG-14713 This is a squashed commit of the patchset (original ids): 959f84d ("logger: import controller-runtime zap as ctrlzap") b8d9cde ("logger: add logLevel to DSCI devFlags") 2d3efe2 ("components, logger: always use controller's logger") 1ef9266 ("components, logger: use one function for ctx creation") - logger: import controller-runtime zap as ctrlzap To avoid patch polluting with the next changes where uber's zap is imported as `zap`. - logger: add logLevel to DSCI devFlags Allow to override global zap log level from DSCI devFlags. Accepts the same values as to `--zap-log-level` command line switch. Use zap.AtomicLevel type to store loglevel. Locally use atomic to store its value. We must store the structure itsel since command line flags parser (ctrlzap.(*levelFlag).Set()) stores the structure there. In its order ctrlzap.addDefault() stores pointers, newOptions() always sets the level to avoid setting it by addDefaults(). Otherwise SetLevel should handle both pointer and the structure as logLevel atomic.Value. It is ok to store structure of zap.AtomicLevel since it itself contains a pointer to the atomic.Int32 and common level value is changed then. The parsing code is modified version of the one from controller-runtime/pkg/log/zap/flag.go. Deprecate DSCI.DevFlags.logmode. - components, logger: always use controller's logger Since the log level is overridable with its own field of devFlags, do not use logmode anymore. It was used to create own logger with own zap backend in case if devFlags exist. Just add name and value to the existing logger instead. Rename NewLoggerWithOptions back to NewLogger since former NewLogger is removed. Change component logger name. "DSC.Component" is redundant. It was usuful when component's logger was created from scratch, but now when it is always based on the reconciler's one, it's clear that it is a part of DSC. - components, logger: use one function for ctx creation Move both logger and component creation to one function. Just a cleanup. Signed-off-by: Yauheni Kaliuta <[email protected]> * secret-generator controller: avoid reporting 'Secret not found' error in reconcile (#1312) * update: remove dsp with v1(tekton)backend related code (#1281) * update: remove dsp with v1(tekton)backend related code - images - tekton rbac - descriptions Signed-off-by: Wen Zhou <[email protected]> Co-authored-by: Humair Khan <[email protected]> --------- Signed-off-by: Wen Zhou <[email protected]> Co-authored-by: Humair Khan <[email protected]> * update: remove two SA which does not seem valid (#1254) Signed-off-by: Wen Zhou <[email protected]> * Update version to 2.20.0 (#1325) Co-authored-by: VaishnaviHire <[email protected]> * fix: wrong type (#1322) - we should watch IngressController from operator.openshift.io, not the k8s ingress - this matched the cache we added in main Signed-off-by: Wen Zhou <[email protected]> * update golangci-lint to v1.61.0 (#1327) * chore(linter): update golangci-lint to v1.61.0 * chore(linter): fix findings * chore: remove duplicated image in modelmesh (#1336) - this only apply for deplenet odh-model-controller - no need to set it in the default map, a waste Signed-off-by: Wen Zhou <[email protected]> * fix: ensure input CAbundle always end with a newline (#1339) * fix: ensure input CAbundle always end with a newline - missing newline will cause problem when inject data into configmap - this happens when user use kustomize with their own CA for DSCI, it set to use |- not | - fix lint Signed-off-by: Wen Zhou <[email protected]> * update: better way to add newline Signed-off-by: Wen Zhou <[email protected]> --------- Signed-off-by: Wen Zhou <[email protected]> * fix: add check to remove old kube-rbac-proxy container in modelregistry operator, fixes RHOAIENG-15328 (#1341) * fix: add check to remove old kube-rbac-proxy container in modelregistry operator, fixes RHOAIENG-15328 Signed-off-by: Dhiraj Bokde <[email protected]> * update: remove check in ModelReg code, add it to upgrade logic Signed-off-by: Wen Zhou <[email protected]> --------- Signed-off-by: Dhiraj Bokde <[email protected]> Signed-off-by: Wen Zhou <[email protected]> Co-authored-by: Wen Zhou <[email protected]> * fix: wrong name of variable and func lead to always return self-managed (#1362) Signed-off-by: Wen Zhou <[email protected]> * fix: add extra check on servicemesh CRD and svc before proceeding create SMCP CR (#1359) * feat: add extra check on servicemesh CRD and service before proceeding create SMCP CR - when we install operator we might have a race condition: - ossm is not ready to server - dsci already created smcp CR - since we do not know the version of smcp to use - ossm webhook can run into conversion problem since no version specify in smcp CR - test: add more negative test - when CRD does not exist - when service does not exist Signed-off-by: Wen Zhou <[email protected]> --------- Signed-off-by: Wen Zhou <[email protected]> * logger: set loglevel from ZAP_LOG_LEVEL environment variable (#1344) Add possibility to set loglevel with the environment. The name is used used already. Command line switch has priority over environment. [1] https://github.com/opendatahub-io/data-science-pipelines-operator/blob/main/config/base/params.env#L13 Signed-off-by: Yauheni Kaliuta <[email protected]> * fix: added a filter to evicted pods when checking for ready status (#1334) * fix: added a filter to evicted pods Signed-off-by: Alessio Pragliola <[email protected]> --------- Signed-off-by: Alessio Pragliola <[email protected]> * Task/NVPE-32: Cleanup NIM integration tech preview resources (#1369) * chore: cleanup nim integration tech preview resources Signed-off-by: Tomer Figenblat <[email protected]> * chore: apply suggestions from code review Co-authored-by: Wen Zhou <[email protected]> * chore: addressed pr reviews, better logging Signed-off-by: Tomer Figenblat <[email protected]> * chore: set nim cleanup for minors 14 and 15 Signed-off-by: Tomer Figenblat <[email protected]> * chore: nim tp cleanup should remove api key secret Signed-off-by: Tomer Figenblat <[email protected]> * chore: addresed pr reviews, rewrite cleanup func Signed-off-by: Tomer Figenblat <[email protected]> --------- Signed-off-by: Tomer Figenblat <[email protected]> Co-authored-by: Wen Zhou <[email protected]> * update: owners (#1376) Signed-off-by: Wen Zhou <[email protected]> * Updated codecov version to 4.6.0 (#1378) * Update version to 2.21.0 (#1379) Co-authored-by: VaishnaviHire <[email protected]> * Dockerfile, Makefile: make CGO_ENABLED configurable (#1382) The commit a107703 ("feat(fips): enable GO_ENABLED in build (#1001)") enabled CGO which makes problems for builders on non-x86 platforms. Make it as an in the Dockerfile keeping default the same (enabled), but make it possible to override with either environment (`export CGO_ENABLED=0`) or make (`make CGO_ENABLED=0 image-build`) Signed-off-by: Yauheni Kaliuta <[email protected]> * Makefile: make USE_LOCAL overridable (#1384) Get USE_LOCAL image build flag from makefile variable to make it overridable with `make USE_LOCAL=true image` Do not allow to get its value for environment be default due to pretty generic name. This is shorter than the old recommendation of overriding IMAGE_BUILD_FLAGS. And since now CGO_ENABLED is also a flag, does not mess up with it. * USE_LOCAL=true uses existing manifests from opt/manifests for the produced image without downloading them with get_all_manifests.sh making it possible to both save time and make local amendments. Signed-off-by: Yauheni Kaliuta <[email protected]> * Dockerfile: merges manifests builder stages to one (#1381) We can combine two build stages into one, as there is no need to always build both images (not done by podman) to only then decide from which one we want to copy manifests to the target image. Instead manifests stage will either copy local manifests or fetches using the script based on USE_LOCAL argument. Move USE_LOCAL and OVERWIRTE_MANIFESTS args under FROM since args have scope of the FROM they are declared in. It requires opt/manifests directory to exist, but since it's a part of git repo, it's fine. Original patch from: Bartosz Majsak <[email protected]> [1] [1] #773 Signed-off-by: Yauheni Kaliuta <[email protected]> * Use incubating for odh-model-controller manifests (#1393) * Use incubating for odh-model-controller manifests * Use kserve v0.14 branch for manifests * add enablement flag for nim (#1330) * add enablement flag for nim * add generated files * add api generated file * revert image for csv * update NIM with its own spec * create nim struct remove from serverless * updated generated files * update nim to default removed * update image to correct var * update to managed as default * add nim-state to params.env * fix NVIDIA typo * resolve conflict * add NIM flag check to model mesh * fix condtional for standard practice nim flag * fix to conditional for standards * update typo for error output nim flag kserve * Fix deploy to dependent path * update doc/manifests references for nim flag * fix image typo csv * update if on nim-state * Remove adelton from the platform owners list. (#1402) * Typo fixes "manifests" (#1409) * doc: remove -e from make invocation examples (#1408) The intention in the examples is to override make variables which is done with setting the variable in the command line, while -e switch does not get argument and changes make to a mode when environment takes precedence over makefile settings[1][2] Short examples: ``` % cat Makefile VAR = var DEFVAR ?= defvar all: @echo "VAR $(VAR), origin $(origin VAR)" @echo "DEFVAR $(DEFVAR), origin $(origin DEFVAR)" % make VAR var, origin file DEFVAR defvar, origin file % make VAR=cmd DEFVAR=cmd VAR cmd, origin command line DEFVAR cmd, origin command line % VAR=env DEFVAR=env make VAR var, origin file DEFVAR env, origin environment % VAR=env DEFVAR=env make -e VAR env, origin environment override DEFVAR env, origin environment % VAR=env DEFVAR=env make -e VAR=cmd DEFVAR=cmd VAR cmd, origin command line DEFVAR cmd, origin command line ``` [1] https://www.gnu.org/software/make/manual/html_node/Values.html [2] https://www.gnu.org/software/make/manual/html_node/Environment.html Signed-off-by: Yauheni Kaliuta <[email protected]> * (fix): Avoid using pull_request_target event trigger. (#1417) Using pull_request_target event trigger results in security vulnerabilities explained here: https://securitylab.github.com/resources/github-actions-preventing-pwn-requests/ Signed-off-by: AJAY JAGANATHAN <[email protected]> * fix: - missing authentications rbca - merge problem - duplicated API definition - lint Signed-off-by: Wen Zhou <[email protected]> --------- Signed-off-by: Yauheni Kaliuta <[email protected]> Signed-off-by: Wen Zhou <[email protected]> Signed-off-by: AJAY JAGANATHAN <[email protected]> Signed-off-by: Dhiraj Bokde <[email protected]> Signed-off-by: Alessio Pragliola <[email protected]> Signed-off-by: Tomer Figenblat <[email protected]> Co-authored-by: Yauheni Kaliuta <[email protected]> Co-authored-by: Wen Zhou <[email protected]> Co-authored-by: Adrián Sanz Gómiz <[email protected]> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: VaishnaviHire <[email protected]> Co-authored-by: Ajay Jaganathan <[email protected]> Co-authored-by: Gerard Ryan <[email protected]> Co-authored-by: Luca Burgazzoli <[email protected]> Co-authored-by: ajaypratap003 <[email protected]> Co-authored-by: Marek Laššák <[email protected]> Co-authored-by: Humair Khan <[email protected]> Co-authored-by: Dhiraj Bokde <[email protected]> Co-authored-by: Alessio Pragliola <[email protected]> Co-authored-by: Tomer Figenblat <[email protected]> Co-authored-by: Hannah DeFazio <[email protected]> Co-authored-by: Marcus Trujillo <[email protected]> Co-authored-by: Jan Pazdziora <[email protected]> Co-authored-by: Christian Zaccaria <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
We can combine two build stages into one, as there is no need to always build both images to only then decide from which one we want to copy manifests to the target image. Instead
manifestsstage will either copy local manifests or fetches using the script based onFETCH_MANIFESTSargument (previously known asUSE_LOCAL).In addition, it opens up
IMAGE_BUILD_FLAGSvariable to be overwritten when executing make targets instead of changing its value in theMakefiledirectly, as it was previously needed.How Has This Been Tested?
Merge criteria: