Do store/expose logs for (at least failed) runs.

[yarikoptic]

Prior ref:

• https://github.com/dandi/dandisets-healthstatus/issues/19

So we do store the logs from each run combined across all assets, e.g.

(venv) dandi@drogon:~/cronlib/dandisets-healthstatus$ ls -lta  results/000003/ | head -n 6
total 4280
drwxr-xr-x 331 dandi dandi 12288 Mar  6 10:36 ..
drwxr-xr-x   2 dandi dandi 86016 Mar  6 10:36 .
-rw-r--r--   1 dandi dandi 16530 Mar  6 10:36 status.yaml
-rw-r--r--   1 dandi dandi 90740 Jan  6 00:34 2024.01.03.12.21.19_matnwb_nwbRead_errors.log
-rw-r--r--   1 dandi dandi  1125 Nov 18 02:13 2023.11.18.00.50.33_matnwb_nwbRead_errors.log

(venv) dandi@drogon:~/cronlib/dandisets-healthstatus$ grep '^Asset' results/000003/2024.01.03.12.21.19_matnwb_nwbRead_errors.log | nl | tail -n 2
    80  Asset: sub-YutaMouse57/sub-YutaMouse57_ses-YutaMouse57-161014_behavior+ecephys.nwb
    81  Asset: sub-YutaMouse57/sub-YutaMouse57_ses-YutaMouse57-161016_behavior+ecephys.nwb

so to access a fail for some sample failing healthcheck, someone needs to

• have access to drogon
• login to it
• find folder
• find most recent log file which would have asset in question listed

which is quite cumbersome. So, ideally we should have them committed to git and pushed directly here on github. The problem is that we can't keep history growing since then repo would grow too large too quickly. But I think we will be ok if we keep only the single state commit and rewrite it. So let's

• create logs/ submodule (https://github.com/dandi/dandisets-healthstatus-logs repo, created now)
• while running a test on a file, create corresponding to the asset path folder (e.g. sub-YutaMouse57/sub-YutaMouse57_ses-YutaMouse57-161014_behavior+ecephys.nwb/
• write under that folder stderr per test, e.g matnwb_nwbRead_errors.txt
• if file ends up empty after run -- remove
• when saving results from a run, commit changes to logs/ --ammend'ing prior commit, or pretty much can recreate that branch from scratch every time, just reusing the tree of prior state
• force push logs/

In the README.md here, per each dandiset row, add there a convenience link to the logs, so row looks like

000019 [logs] | 30 passed, 1 failed, 0 timed out | 0 passed, 31 failed, 0 timed out | —

Ideally, if it is feasible, if we could as a comment in results/ yaml files urls to logs, like

tests:
- assets_nok:
  - sub-EC2/sub-EC2_ses-EC2-B1.nwb  # https://github.com/dandi/dandisets-healthstatus-logs/blob/main/000019/sub-EC2/sub-EC2_ses-EC2-B1.nwb/pynwb_open_load_ns_errors.txt
  assets_ok:
  - sub-EC2/sub-EC2_ses-EC2-B105.nwb
  name: pynwb_open_load_ns

that would have been great, but I have not looked how feasible it is yet.

[yarikoptic]

ping on this

[yarikoptic]

@rly please sign up to this issue to know when you can get to the logs seamlessly