Goal
make down stops the stack and keeps the volumes — correct for a real
deployment, wrong for someone evaluating who wants a clean slate. Neither
Makefile nor deploy/community/README.md says how to actually discard the
data and start again, so the first thing an evaluator wants to do after a failed
experiment is undocumented.
Where
deploy/community/README.md — the Quick start and Updates and scaling
sections are the natural neighbours
Makefile — optional; if you add a target, see the warning below
What to do
Document the reset: how to remove the volumes, what is destroyed with them
(database, uploads, figures, the erasure journal), and that this is not an
upgrade path.
If you add a make target for it, it must be hard to run by accident — a bare
make reset that silently drops a production database is a worse outcome than
no target at all. Require an explicit confirmation argument, and say in the
## description that it destroys data. If you would rather leave it as
documented commands only, that is a legitimate answer; say so in the pull
request.
How to verify
Run the commands you wrote, from a stack you started yourself, and confirm the
next make up really does start from an empty database. Say in the pull request
that you did.
Acceptance criteria
Goal
make downstops the stack and keeps the volumes — correct for a realdeployment, wrong for someone evaluating who wants a clean slate. Neither
Makefilenordeploy/community/README.mdsays how to actually discard thedata and start again, so the first thing an evaluator wants to do after a failed
experiment is undocumented.
Where
deploy/community/README.md— theQuick startandUpdates and scalingsections are the natural neighbours
Makefile— optional; if you add a target, see the warning belowWhat to do
Document the reset: how to remove the volumes, what is destroyed with them
(database, uploads, figures, the erasure journal), and that this is not an
upgrade path.
If you add a
maketarget for it, it must be hard to run by accident — a baremake resetthat silently drops a production database is a worse outcome thanno target at all. Require an explicit confirmation argument, and say in the
##description that it destroys data. If you would rather leave it asdocumented commands only, that is a legitimate answer; say so in the pull
request.
How to verify
Run the commands you wrote, from a stack you started yourself, and confirm the
next
make upreally does start from an empty database. Say in the pull requestthat you did.
Acceptance criteria
make downand frommake updateis clearbash deploy/community/test.shpasses if you touched any script