Reference-count flightCall so the fan-out context ends with the last
caller waiting on it, keeping cfg.Timeout as the upper bound. A leader
leaving with a follower still parked no longer disturbs the flight, and
a solo requester disconnecting releases the upstream sockets at once
instead of holding them for the whole timeout.
Return errFlightAbandoned from Do rather than inferring the abandon path
from the request context's sentinel, and read Age off the server's clock
so it matches the timestamp the cache stored.
A stale fallback is byte-identical to a fresh response, so a client has
no way to tell it is holding data pdbmux served only because every
backend was down; the sole signal is a log line and a /healthz counter.
Set X-Cache to hit, miss or stale and Age to whole seconds since the
served copy was stored on every response from a cached path. Neither
header is emitted by OpenVoxDB, so nothing upstream is shadowed.
A single flight is built by whichever request arrived first, but every
request on that key waits for it. Running the fan-out on the leader's
cancelable request context hands the leader's disconnect to followers
whose own connections are healthy: they get 502 all backends failed.
Waiters also parked on a WaitGroup, so a follower whose own client went
away stayed blocked until the leader finished.
Run the flight on a context detached from the leader's request and
bounded by the configured timeout, and pass that context into build so
the fan-out uses it. Give Do a context so a waiter can abandon a flight
it no longer needs; the leader ignores it and always runs fn to
completion, keeping the cache warm for the others. A caller that
abandons on its own cancellation writes no response.
A busy Puppetboard re-fans-out the same /facts query every few seconds, and a
502 is worse than 30-second-old facts when every PuppetDB is unreachable.
- Add a `Cache` interface (get reports fresh/stale/miss, put, stats) keyed on
`<path>?<params>` with keys and repeated values sorted, plus a no-op default
so uncached paths behave exactly as before.
- Route serveMerged/serveUnion/serveSummed through `serveCached`, so the
reports cache drops in at `cacheFor` without touching a handler.
- Back /facts and /nodes with a byte-bounded LRU: `facts_ttl` (default 30s,
clamped to a 30s cap) and `facts_cache_bytes` (default 64 MiB); expired
entries are kept and served only when every backend fails.
- Single-flight identical keys so N concurrent requests cause one fan-out.
- Surface `cache` state and `serving_stale` in /healthz and the cache settings
in `config show`.
- drop upstream facts of the configured name whenever the feature is enabled,
independent of the per-query injection gate, and log the drop once per request
- walk the whole AST for a nested extract and skip injection when one is found
outside an in subquery
- skip the environment scan on /facts when nothing is injected
- document the override rule and that PQL-syntax queries never get the fact
## Why
With several PuppetDBs behind one endpoint, consumers cannot tell which backend a node's data came from.
## How
- Add a synthetic `pdbmux_source` fact per certname on `/facts`, valued with the backend that won the facts merge, and stamp the same key on merged `/nodes` records.
- Emit all four fact keys including `environment`, which clients index directly.
- Skip injection for top-level `extract` queries, so `count()` and other aggregates keep the backends' own numbers, and for `/facts` queries constraining `name`; a `name` filter inside an `in` subquery still injects.
- Replace, never duplicate, an upstream fact of the configured name.
- Configure via `source_fact` / `source_fact_enabled` (`PDBMUX_SOURCE_FACT`, `PDBMUX_SOURCE_FACT_ENABLED`), defaulting to `pdbmux_source` enabled.
## Why
Puppetboard 7.0.1 cannot run against pdbmux: it exits at import when
/pdb/meta/v1/version 404s, and its landing page, metrics and radiator views
404 on the Jolokia surface.
## How
- Serve /pdb/meta/v1/version, reporting the lowest version any backend runs,
and /pdb/meta/v1/server-time from the first reachable backend.
- Merge the Jolokia surface (/metrics/v2/read, /metrics/v2/list,
/metrics/v1/mbeans): objects union, numeric attributes sum by default, and
Min/Max/Uptime/StartTime plus the distribution stats take a bound or a mean.
- Route /nodes extract-count queries to the summing path ahead of the certname
merge, and give /resources aggregates the same path.
- Document the endpoints and merge semantics in the README.
- Cover version disagreement, metric rules, escaped MBean names, count summing
and the non-aggregate /nodes merge with httptest backends.
An extract/count()/group_by query returns synthetic rows with no report
hash, which reportKey fell back to keying by verbatim raw bytes. Two
backends emitting a byte-identical aggregate row (e.g.
{"status":"changed","count":1}) therefore collapsed into one, silently
undercounting the merged result and contradicting the documented
guarantee that no backend's rows are dropped.
Give the mergeUnion key func an ok return: false means the record has no
dedupe identity and is always kept. reportKey returns ok=false for
hash-less rows; hash-keyed report dedupe and event verbatim-identity
dedupe are unchanged.
Add TestMergeUnion_IdenticalHashlessRowsAreNotCollapsed covering the
collision case, and reword the README line to say aggregate rows pass
through even when byte-identical.
Reports are immutable history, so a node that migrated has reports in the
old PuppetDB and the new one; serve the union rather than picking a single
owning backend as /facts does.
Re-apply order_by/limit/offset over the merged set and sum X-Records, since
each backend only orders and pages its own slice.
Split out from node-lookup PR #17 into its own repo. pdbmux presents a
single merged PuppetDB v4 query surface over the old (Consul) and new
(k8s) PuppetDBs during the VM to k8s migration, and is deployed in-cluster
via argocd-apps as a container image.
unkin-agent
benvin