artifact-store-tools: say an image a definition pins is named by a definition, under any name the store has had (issue 323)
mesh/merge-gate pass: builds artifact-store-tools → novox; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/repo-check pass: its merge-check.sh passed, each part in its toolchain: merge-check.sh (go), merge-check-typescript.sh (typescript)
mesh/delivery delivered

No record names an image the mesh did not make, so one a definition pins under an
older name of the store read as unrecorded, the same word as an image nothing
uses. The tool now reads every definition through the catalogue and says
named-by-a-definition with the module. The store's older names are the mesh's
setting older-store-names, since a catalogue names no installation.
This commit is contained in:
jochen
2026-10-12 14:47:06 +02:00
parent 2772120758
commit d745a4d31e
9 changed files with 777 additions and 33 deletions
+22 -4
View File
@@ -1,7 +1,8 @@
# artifact-store-tools
Tools that say what the mesh's artifact store holds, and what the controller's records say of it
(novox/hq ADR 0251 §1–3, to-be 51). It declares no resources and changes nothing on its machine.
(novox/hq ADR 0251 §1–3, to-be 51). It changes nothing on its machine; its only resources are its state
directory and the file the mesh writes there from its setting `older-store-names`.
## Why a module beside the store
@@ -18,7 +19,7 @@ on that machine.
|---|---|---|
| `artifact_store_repositories` | r | every repository: tags and what each names, how many manifests (tagged or not), size |
| `artifact_store_usage` | r | the store's size, the largest repositories, bytes shared between repositories, bytes no manifest marks (what the nightly collector frees next) |
| `artifact_store_references` | r | what the controller's records say of each manifest, counted and sized by state and repository; each manifest listed when one repository is asked; what the records keep that the store does not hold |
| `artifact_store_references` | r | what the controller's records and the catalogue's definitions say of each manifest, counted and sized by state and repository; each manifest listed when one repository is asked; what the records keep that the store does not hold |
| `artifact_store_collect` | a | a dry run unless `dry_run` is false: what the controller would let go of, and the bytes the nightly collector would free then and now. A real run needs `why` and asks the controller's `collect` |
### How the store is read
@@ -43,6 +44,7 @@ it lists. The store's collector removes every blob no manifest marks, so these n
| state | what | removable |
|---|---|---|
| `kept` | a definition names it, or one of the five most recent builds of a module the mesh holds; `why` says which | no |
| `named-by-a-definition` | no record names it, and a definition in the catalogue pins it by digest, under the store's current name or an older one; `why` names the module (hq issue 323) | no |
| `holder-of-kept-archive` | the manifest that keeps a kept archive's blob (hq issue 253) | no |
| `eligible` | the mesh made it and keeps it for no reason | through the controller's `collect` |
| `holder-of-eligible-archive` | the manifest that keeps an eligible archive's blob | with its archive |
@@ -56,6 +58,20 @@ its own right; the controller lets an index go with the platform manifests only
counted as the index's. Two indexes naming one platform give it the more keeping state. An index no
record names leaves its platform manifests unrecorded with it.
**What the definitions pin** (hq issue 323). A definition may pin an image of the store the mesh did not
make: pushed before the mesh built it, and pinned under the store's current name or a name it had before.
No record of the controller names it, yet a machine runs it, so `artifact_store_references` reads every
definition the catalogue holds (`mesh-catalog.catalog_modules`, then `catalog_module` for each, eight at
a time and within twenty seconds) and finds every `<name>/<repository>@sha256:…` in any of its strings,
under each of the store's names: `artifact-store://`, the store's address as the controller answers it,
and the names it had before. Those are the mesh's setting `older-store-names` (space-separated, `none` by
default, set with the controller's `settings`), which the mesh writes to the module's `store-names` file: a
catalogue names no installation. When they cannot be read, the answer's first line says so. A manifest no record
names that a definition pins is `named-by-a-definition`; one the records name keeps the records' state. A
definition read at its commit is kept, so the next call asks only for the list. Definitions that could
not be read are listed under `definitions.unread`, and the answer's first line says that what they pin
reads as unrecorded until they are.
**Unrecorded copies of upstream images** are listed in `unrecorded_mirrors`: each manifest at the top of
a repository only the build's mirror writes (`upstream/<host>/<path>`, or a module's `<module>/on-<argument>`
from before hq ADR 0257) that no record names, with what it and its platforms mark alone. They are the
@@ -88,6 +104,8 @@ with slashes, a cut listing refused), manifests read with one absent or unreadab
digest never becoming a path outside the blobs, the mark (an index, a shared blob, a blob nothing marks),
every state, what the records keep that the store lacks, bytes freed counting a blob shared with a kept
manifest as kept, a dry run never confirming, a real run without `why` refused before anything is asked,
the fall-back when the references let go of cannot be carried, escalation through `sudo -n`, and that the
tools served are exactly the manifest's `tools` and its `invokes` exactly the two controller verbs.
the fall-back when the references let go of cannot be carried, a manifest a definition pins under the
store's current or an older name, a definition not read said first, escalation through `sudo -n`, and that the
tools served are exactly the manifest's `tools` and its `invokes` exactly the two controller verbs and the
catalogue's two tools.
Any command other than the two read-only scripts fails the test.