feat: Mirror Core expected inventory into Flow tables#2440
Open
kunzhao-nv wants to merge 7 commits into
Open
Conversation
Contributor
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Enterprise Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
🔐 TruffleHog Secret Scan✅ No secrets or credentials found! Your code has been scanned for 700+ types of secrets and credentials. All clear! 🎉 🕐 Last updated: 2026-06-11 00:18:56 UTC | Commit: 89473f2 |
kunzhao-nv
force-pushed
the
feat/flow-expected-inventory-sync
branch
from
89473f2 to
f6a5afb
Compare
Signed-off-by: Kun Zhao <kunzhao@nvidia.com>
Nullable + partial unique index so legacy ingestion rows coexist until the mirror adopts them. Signed-off-by: Kun Zhao <kunzhao@nvidia.com>
kunzhao-nv
changed the title
🔍 Container Scan Summary
Per-CVE detail lives in the per-service |
Runs at the top of each inventory cycle. Adopts legacy rows by (manufacturer, serial), soft-deletes ones Core no longer reports, resurrects soft-deleted rows when Core re-reports them. Mirrors racks whose Core rack_id is unset (external_id stays NULL) and GCs stale soft-deleted tombstones occupying a name that Core is reusing so the INSERT/UPDATE doesn't collide on the full unique rack_name_idx. RPC errors and empty Core responses skip the delete phase. Signed-off-by: Kun Zhao <kunzhao@nvidia.com>
Match by (manufacturer, serial_number) — same unique key Flow's ingestion path already enforces — and reconcile the host BMC alongside. Non-host BMC rows (DPU, etc.) are left strictly alone: Core's ExpectedMachine doesn't describe them, so touching them from here would hard-delete the DPU BMC whenever the bun preload order put it at BMCs[0]. RackID is resolved per cycle from rack.external_id so a machine's RackID="a12" lands on the matching Flow rack UUID. Soft-delete + resurrect on transient absence, same shape as the rack mirror. firmware_version and BMC credentials are not mirrored: the former is owned by the runtime sync's BMC inventory read, the latter live in Vault after site-explorer's password rotation. Each insert / update (with field-by-field deltas) / delete is logged at INFO so the operator can audit what Core drove this cycle. Signed-off-by: Kun Zhao <kunzhao@nvidia.com>
…ync_* files Move the rack-specific code out of expected_mirror.go into expected_mirror_rack.go. Move each per-type actual sync (machine / switch / powershelf) and its tests out of inventory.go into actual_sync_*.go; cross-type helpers land in actual_sync.go. inventory.go shrinks to the runInventoryOne orchestrator. Signed-off-by: Kun Zhao <kunzhao@nvidia.com>
- Resolve-rack-id failure now mirrors the component with NULL rack_id instead of soft-deleting it. Core is still reporting the component; losing its UUID because the rack mirror dropped one cycle ahead is wrong. - BMC insert paths first hard-delete any orphan owning the same MAC. The bmc table's PK is the MAC alone so a chassis-to-chassis BMC transplant without orphan cleanup would otherwise roll back the whole type's mirror tx. - planBMCReconciliation now finds every type='Host' BMC and hard-deletes the extras. Core's authoritative view is exactly one host BMC per component; the mirror enforces it. - parseLabelInt distinguishes "Core omitted the label" (0, ok=true) from "Core sent a non-integer" (0, ok=false). Malformed labels mark the field for preservation so UPDATE keeps Flow's existing value instead of clobbering with 0; INSERT still falls back to 0 with a warn since there's nothing to preserve. - Duplicate (manufacturer, serial) specs from Core get a warn so the Cloud REST bug is visible; last write wins. - Component and rack mirror reset success-side counters on tx rollback so the summary log reflects what actually landed. Signed-off-by: Kun Zhao <kunzhao@nvidia.com>
kunzhao-nv
force-pushed
the
feat/flow-expected-inventory-sync
branch
from
f6a5afb to
f03a6e4
Compare
|
🌿 Preview your docs: https://nvidia-preview-pull-request-2440.docs.buildwithfern.com/infra-controller |
…TORY_SYNC_ENABLED Default is off so deploying the binary doesn't auto-activate writes to rack / component from the new mirror path. Operators flip the env to a strconv.ParseBool-truthy value per cluster to opt in. The gate is read once at job construction; runInventoryOne skips syncExpectedFromCore entirely when off. Signed-off-by: Kun Zhao <kunzhao@nvidia.com>
kunzhao-nv
force-pushed
the
feat/flow-expected-inventory-sync
branch
from
f03a6e4 to
9a79c7f
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
rack,componentandbmctables on every inventory cycle, before drift detection runs. Gated byFLOW_EXPECTED_INVENTORY_SYNC_ENABLED, default off — deploying this PR is a no-op until the env is flipped per-cluster. Theinventorysyncpackage is split intoexpected_mirror_*+actual_sync_*files (separate refactor commit) so the new code is easy to navigate.rack_id(newrack.external_idcolumn + migration); component by(manufacturer, serial_number). Soft-deleted rows resurrect on Core re-report so UUIDs stay stable; rack-name tombstones are GC'd in-tx for index-collision safety.type='Host'rows — DPU BMCs are left to the ingestion path. Extra host BMCs are hard-deleted to enforce Core's "exactly one host" view; MAC-PK conflicts evict the orphan in-tx so the cycle doesn't abort.Type of Change
Breaking Changes
Testing