Repository navigation
fix(fleet): summarize reserved and observed GPU runs on one line - #523
Merged
Merged
Conversation
dmlguq456
force-pushed
the
fix/fleet-gpu-oneline-1011
branch
from
October 11, 2026 09:10
5806366 to
f5829dd
Compare
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fleet's folded compute panel added a reservation row for every GPU run, duplicating the process count and stretching the default screen. Keep one row per GPU and append observed process counts by session tag, sorted by count; narrow screens shorten whole entries to
+Nwhile preserving metric columns.Only reservation/process differences add
reserved+Norunreserved+N(shortened when necessary). Counts compare full session identities, use current shared session tags, and avoid treating failed queries as empty samples. The existingpexpansion retains reservation owners, tasks, timestamps, and process commands, including reservation-only GPUs. All harnesses use the same read-only renderer.Validation: 166 tests passed: 7 focused summary/width/mismatch/expansion tests, 125 existing GPU collector, owner-link, NVML-fallback, unregistered-process and gauge tests, and 34 screen-bundle tests. Independent review passed after correcting a dual-mismatch warning overflow with a 10-character hostname at 80 columns. Updated three screen fixtures for the added summary, and verified missing reservation payloads remain unknown. Captured actual before/after
--once --allscreens at 200 and 80 columns. GPU processes and reservation storage were only read.