methodology
Explore the methodology and documentation practices to better move this endeavor along.
2025-02-03
In the most recent test, it has not been possible to provoke corruption using the S/W versions from the original issue after running for over 1100 syncoid
invocations. At this point I am stepping back to try for some improvements.
- The first priority is to explore the test methodology to see if results can be achieved in less time.
- The second priority is to standardize the testing procedure and recording policies to provide more reliable results in a manner that is easier to view (IOW, It's hard find the details of stuff I did weeks ago!)
- To facilitate the previous, The notes will be organized such that they can be processed using
mkdocs
.
Organization
"Test cases" will be listed in the README in .../provoke_ZFS_corruption/docs/tests
and will include
- date started
- date completed
- results
- link to test_setup description
- ZFS version or commit hash (and whether or not from backports, release, etc.)
- OS variant (e.g. Buster, Bullseye, Bookworm)
- kernel version (and whether or not from backports.)
Each "test case" will have a set of files in it's own directory named starting with the date in ISO-8601 format (e.g. 2025-02-03_test
)
setup.md
Commands required to set up the test and kick it off along with any relevant results. The markdown title will include the date string and "setup" (e.g.# 2025-02-03 Setup
, first line in the file.)results.md
Copies of text from the screen for anything that might be worthy of later review.results.md
andsetup.md
sho0uld link to each other.[anything else useful]
as the situation warrants.