Repository navigation
D240 — a sealed run stays under the user's kernel limits, which no folder of its own isolates - #438
Open
napkinstack-agent[bot] wants to merge 1 commit into
Open
napkinstack-agent[bot] wants to merge 1 commit into
napkinstack-agent[bot] wants to merge 1 commit into
Conversation
…lder of its own isolates Downstream, after D233: one exhausted per-user limit (inotify instances) cost several end-to-end runs in a week. The sealed table in tests.md gains a sixth row — file watchers, open files, processes — with its check. Documentation only; nothing green turns red. Refs #419. Agent-Session: nstack-os-d240 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why
After D233, a downstream project read its sealed end-to-end runs against the new table. One exhausted per-user kernel limit, inotify instances, still cost it several runs in one week. No row of the table covers that: a folder of its own run, a process group or port 0 isolate nothing of a limit the kernel charges per user. Any stack whose tools watch files (bundlers, dev servers, test watchers) or start many processes meets it.
What
skeleton/.nstack/playbooks/tests.md§1: the sealed table gains a sixth row, per-user kernel limits — file watchers (inotify instances and watches) and process count. The check: the run closes what it opens, as many runs at once stay underfs.inotify.max_user_instances,fs.inotify.max_user_watchesandulimit -u(threads included), and a limit run out fails at once, naming it.ulimit -n) is left out on purpose: it is per process, and runs never share it.sealed()docstring give the same list as the table.Documentation and a message. No rule added; nothing green turns red.
Evidence
test_a_sealed_run_names_everything_it_must_not_share_and_how_to_check_itwas extended first and seen red (5 rows, no kernel limits); it is green now. It also pins both inotify keys andulimit -u, refusesulimit -n, and binds M12's message to the table.uv run bash platform/tests/run.sh: Platform tests OK.ulimit -nwrongly named as shared,max_user_watchesmissing from the check, M12's message drifting from the table.Refs #419.
🤖 Generated with Claude Code