Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Search][Index Management] Removing Model deployment from Kibana #198409

Merged
merged 10 commits into from
Nov 6, 2024
Merged

[Search][Index Management] Removing Model deployment from Kibana #198409

Show file tree
Hide file tree
Changes from 9 commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
518a96e
Disabling try again button while model is still deploying
Samiul-TheSoccerFan Oct 30, 2024
ab211a0
Merge branch 'main' into fix-model-redeploying-issue
elasticmachine Oct 30, 2024
9910441
Merge branch 'main' into fix-model-redeploying-issue
elasticmachine Oct 30, 2024
4d820d5
Merge branch 'main' into fix-model-redeploying-issue
elasticmachine Oct 30, 2024
7ceb873
revert back the disable logic for try again button
Samiul-TheSoccerFan Oct 31, 2024
df1051a
Adding restriction to avoid redeploying models from kibana
Samiul-TheSoccerFan Oct 31, 2024
2128a66
Removing model deployment from kibana - index management
Samiul-TheSoccerFan Nov 6, 2024
278311a
Merge branch 'main' into fix-model-redeploying-issue
elasticmachine Nov 6, 2024
73aa920
Merge branch 'main' into fix-model-redeploying-issue
elasticmachine Nov 6, 2024
7ff2649
Reverting back to deployment id to point to inference id instead of m…
Samiul-TheSoccerFan Nov 6, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 1 addition & 39 deletions ...lic/application/sections/home/index_list/details_page/trained_models_deployment_modal.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,10 +26,8 @@ import React from 'react';
import { EuiLink } from '@elastic/eui';
import { useEffect, useMemo, useState } from 'react';
import { i18n } from '@kbn/i18n';
import { ModelIdMapEntry } from '../../../../components/mappings_editor/components/document_fields/fields';
import { isSemanticTextField } from '../../../../components/mappings_editor/lib/utils';
import { deNormalize } from '../../../../components/mappings_editor/lib';
import { useMLModelNotificationToasts } from '../../../../../hooks/use_ml_model_status_toasts';
import { useMappingsState } from '../../../../components/mappings_editor/mappings_state_context';
import { useAppContext } from '../../../../app_context';

Expand All @@ -55,15 +53,11 @@ export function TrainedModelsDeploymentModal({
}: TrainedModelsDeploymentModalProps) {
const modalTitleId = useGeneratedHtmlId();
const { fields, inferenceToModelIdMap } = useMappingsState();
const {
plugins: { ml },
url,
} = useAppContext();
const { url } = useAppContext();
const [isModalVisible, setIsModalVisible] = useState<boolean>(false);
const closeModal = () => setIsModalVisible(false);
const [mlManagementPageUrl, setMlManagementPageUrl] = useState<string>('');
const [allowForceSaveMappings, setAllowForceSaveMappings] = useState<boolean>(false);
const { showErrorToasts, showSuccessfullyDeployedToast } = useMLModelNotificationToasts();

useEffect(() => {
const mlLocator = url?.locators.get(ML_APP_LOCATOR);
Expand All @@ -86,25 +80,6 @@ export function TrainedModelsDeploymentModal({

const [pendingDeployments, setPendingDeployments] = useState<string[]>([]);

const startModelAllocation = async (entry: ModelIdMapEntry & { inferenceId: string }) => {
try {
await ml?.mlApi?.trainedModels.startModelAllocation(entry.trainedModelId, {
number_of_allocations: 1,
threads_per_allocation: 1,
priority: 'normal',
deployment_id: entry.inferenceId,
});
showSuccessfullyDeployedToast(entry.trainedModelId);
} catch (error) {
setErrorsInTrainedModelDeployment((previousState) => ({
...previousState,
[entry.inferenceId]: error.message,
}));
showErrorToasts(error);
setIsModalVisible(true);
}
};

useEffect(() => {
const models = inferenceIdsInPendingList.map((inferenceId) =>
inferenceToModelIdMap?.[inferenceId]
Expand All @@ -114,18 +89,6 @@ export function TrainedModelsDeploymentModal({
}
: undefined
); // filter out third-party models
for (const model of models) {
if (
model?.trainedModelId &&
model.isDeployable &&
!model.isDownloading &&
!model.isDeployed
) {
// Sometimes the model gets stuck in a ready to deploy state, so we need to trigger deployment manually
// This is currently the only way to surface a specific error message to the user
startModelAllocation(model);
}
}
const allPendingDeployments = models
.map((model) => {
return model?.trainedModelId && !model?.isDeployed ? model?.inferenceId : '';
Expand All @@ -135,7 +98,6 @@ export function TrainedModelsDeploymentModal({
(deployment, index) => allPendingDeployments.indexOf(deployment) === index
);
setPendingDeployments(uniqueDeployments);
// eslint-disable-next-line react-hooks/exhaustive-deps
}, [inferenceIdsInPendingList, inferenceToModelIdMap]);

const erroredDeployments = pendingDeployments.filter(
Expand Down
6 changes: 3 additions & 3 deletions x-pack/plugins/index_management/public/hooks/use_details_page_mappings_model_management.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,9 +34,9 @@ const getCustomInferenceIdMap = (
? {
trainedModelId: model.service_settings.model_id,
isDeployable: model.service === Service.elser || model.service === Service.elasticsearch,
isDeployed: modelStatsById[model.inference_id]?.state === 'started',
isDeployed: modelStatsById[model.service_settings.model_id]?.state === 'started',
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think this is correct. We should be looking at the inference ID, because we don't care about the model in general, we care about the deployment of the model that is specific to the inference ID.

isDownloading: Boolean(downloadStates[model.service_settings.model_id]),
modelStats: modelStatsById[model.inference_id],
modelStats: modelStatsById[model.service_settings.model_id],
}
: {
trainedModelId: '',
Expand Down Expand Up @@ -104,7 +104,7 @@ export const useDetailsPageMappingsModelManagement = () => {
Record<string, TrainedModelStat['deployment_stats'] | undefined>
>((acc, { model_id: modelId, deployment_stats: stats }) => {
if (modelId && stats) {
acc[stats.deployment_id] = stats;
acc[modelId] = stats;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think this is correct. Semantic text models should be referenced by their deployment ID, not their model ID, because they only reference the model ID with the deployment ID that is identical to the inference endpoint name.

}
return acc;
}, {}) || {};
Expand Down