Skip to content

Keep tile module logs after their temp run dirs are removed - #948

Open
cailmdaley wants to merge 1 commit into
developfrom
fix/keep-tile-module-logs
Open

cailmdaley wants to merge 1 commit into
developfrom
fix/keep-tile-module-logs

Conversation

@cailmdaley

Copy link
Copy Markdown
Contributor

Keeps the ShapePipe module logs of tile_vignets and the ngmix chunks after their run directories are deleted.

tile_ngmix's chunk run directories are temp(), so their logs were deleted after tile_merge_cats. Those logs hold the per-chunk "epoch cuts:" summary and the per-object failures. They were lost under both launch modes, even with clean_tiles: false. This is how we couldn't diagnose the 23 objects lost under des_y6 weighting.

Each of these jobs now copies its logs to <tile>/logs/modules/<run>/ before it exits, whatever its exit status. A failed copy only warns. The cost is about 180 KB per chunk.

Test: test_module_logs_outlive_their_run_dirs fails without the copy.

Claude Opus 5.5 on behalf of Cail

🤖 Generated with Claude Code

https://claude.ai/code/session_012yfkBvkzPv33iJxXZ6jAxd

A chunk's ShapePipe run dir is temp(), removed once tile_merge_cats has
read it, and tile_vignets' run dir lives in the node-local store, so both
runs' logs went with their dirs: the ngmix process logs that carry each
chunk's `epoch cuts:` line and every per-object "ngmix failed" message
were gone after any campaign, classic or single-allocation. Each rule now
copies its run's logs/ files to $SP_RUN/logs/modules/<run name>/ before
it exits, whatever its rc, without changing that rc.

The shell change moves the params pin in tile_vignets and tile_ngmix, so
land it at a campaign boundary.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01D8wdTpGpBtyK6uHwU7qWQW

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant