docsy(v2): set GPU limits in cluster resource config#1212
Open
ppiegaze wants to merge 1 commit into
Open
Conversation
The resource-management guide documented project-domain quotas for CPU and memory (projectQuotaCpu / projectQuotaMemory via `uctl update cluster-resource-attribute`) but omitted GPU entirely, leaving DOC-306's question unanswered. v2/Union has no standard project-level GPU quota key (the projectQuota* keys are free-form ResourceQuota template variables covering CPU/memory). GPU limits are enforced through the settings system via `task_resource.max.gpu` (hard per-task max) and `task_resource.min.gpu` (default request), set with `flyte edit settings`. Add a "GPU limits" subsection stating the confirmed keys and cross-linking Settings. Refs: DOC-306 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Peeter Piegaze <1153481+ppiegaze@users.noreply.github.com>
GHA build & deploy previewBuilt by
Updated automatically on every push. |
ppiegaze
requested review from
cosmicBboy,
kumare3 and
samhita-alla
as code owners
Contributor
There was a problem hiding this comment.
Pull request overview
Adds documentation to the Resource management guide explaining how to configure GPU limits for domains/projects in Union via the hierarchical Settings system (rather than projectQuota* ClusterResourceAttribute variables).
Changes:
- Adds a new GPU limits subsection clarifying there is no standard
projectQuota*GPU key. - Documents the relevant Settings keys (
task_resource.max.gpu,task_resource.min.gpu) and how to edit them withflyte edit settings. - Notes the semantic difference between namespace-wide quotas (CPU/memory) and per-task GPU limits, with a cross-link to the Settings guide.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
This file contains hidden or 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
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.
What
Adds a GPU limits subsection to the Resource management guide (
user-guide/project-patterns/resource-management), answering DOC-306 — "how does an admin set GPU limits in the cluster resource config?"Why / grounding (code-is-authoritative)
The page documented project-domain quotas for CPU and memory only (
projectQuotaCpu/projectQuotaMemoryviauctl update cluster-resource-attribute), with GPU absent. DOC-306 (and its Jira original, DOCS-362) asked specifically how to raise the GPU limit, guessing at aprojectQuotaNvidiaGpukey.Verified against live docs + repo content + source:
projectQuota*attributes are free-form ResourceQuota template variables — the shipped/standard template wires onlylimits.cpu/limits.memory(deployment/flyte-plugins/kubernetes-plugins.mdtemplate;deployment/flyte-configuration/customizable_resources.md— "free-form strings"). SoprojectQuotaNvidiaGpuis not a real, fixed key.task_resource.max.gpu— maximum GPU per task pod, enforced as a hard limit that a per-taskflyte.Resourcesrequest cannot exceed.task_resource.min.gpu— default GPU request when a task omits one.flyte edit settings --domain <d>(or--project). Documented onuser-guide/core-concepts/settingsand confirmed against the live docs (.../core-concepts/settings/page.md).The new subsection states the confirmed keys, notes these are per-task limits rather than a namespace aggregate, and cross-links Settings. Scope is deliberately narrow (the platform/domain/project GPU cap) — distinct from the per-task
Resources(gpu=)API (DOC-1102 / #1197).Hold status
Held draft — the mechanism ships today (already live in the Settings docs), so this is publish-ready on human review/merge, not blocked on an unshipped feature.
Refs: DOC-306