fix(v1.1.7-dead-letter): wire dispatcher → list_matching_dead_letter

dead_letter triggers have been registerable since v1.1.1 but their
handlers never fired: dispatcher::handle_failure wrote the dead_letters
row and stopped — list_matching_dead_letter had no production caller.
Any deploy v1.1.1–v1.1.6 with dead_letter triggers had silently
non-functional handlers.

The fix: after the dead-letter row is inserted on retry exhaustion, fan
out to matching dead_letter triggers (filtered by source / originating
trigger_id / script_id) and enqueue one outbox row per match carrying a
real-shape TriggerEvent::DeadLetter (the §6 brief field names were stale
— used the actual variant: dead_letter_id, original: Box<TriggerEvent>,
attempts, last_error, trigger_id, script_id, first/last_attempt_at).
The recursion-stop (a handler's own failure isn't re-dead-lettered)
is upheld by the existing is_dead_letter_handler short-circuit.

Tests (DB-gated): handler actually fires with the nested original event;
existing row-create test now also asserts handler-fire; source_filter
excludes non-matching; failing handler does not recurse.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
MechaCat02
2026-06-04 22:30:25 +02:00
parent 1f78937dd2
commit 02335a8132
2 changed files with 256 additions and 32 deletions

View File

@@ -23,19 +23,19 @@
use std::sync::Arc;
use std::time::Duration;
use chrono::Utc;
use chrono::{DateTime, Utc};
use picloud_executor_core::{ExecError, ExecRequest, ExecResponse, InvocationType};
use picloud_orchestrator_core::{ExecutionGate, ExecutorClient};
use picloud_shared::{
ExecResponseSummary, ExecutionId, HttpDispatchPayload, InboxDeliveryOutcome, InboxFailureKind,
InboxResolver, InboxResult, RequestId, ScriptId, ScriptSandbox, TriggerEvent,
DeadLetterId, ExecResponseSummary, ExecutionId, HttpDispatchPayload, InboxDeliveryOutcome,
InboxFailureKind, InboxResolver, InboxResult, RequestId, ScriptId, ScriptSandbox, TriggerEvent,
};
use rand::Rng;
use uuid::Uuid;
use crate::abandoned_repo::{AbandonedRepo, NewAbandonedExecution};
use crate::dead_letter_repo::{DeadLetterRepo, NewDeadLetter};
use crate::outbox_repo::{OutboxRepo, OutboxRow, OutboxSourceKind};
use crate::outbox_repo::{NewOutboxRow, OutboxRepo, OutboxRow, OutboxSourceKind};
use crate::principal_resolver::PrincipalResolver;
use crate::repo::ScriptRepository;
use crate::trigger_config::{BackoffShape, TriggerConfig};
@@ -463,12 +463,12 @@ impl Dispatcher {
// Exhausted retries → dead-letter.
let (op, source) = describe_event(&row.payload);
let now = Utc::now();
if let Err(e) = self
let dl_id = match self
.dead_letters
.insert(NewDeadLetter {
app_id: row.app_id,
original_event_id: row.id,
source,
source: source.clone(),
op,
trigger_id: row.trigger_id,
script_id: Some(resolved.script_id),
@@ -480,8 +480,26 @@ impl Dispatcher {
})
.await
{
tracing::error!(?e, "failed to write dead-letter row");
Ok(id) => Some(id),
Err(e) => {
tracing::error!(?e, "failed to write dead-letter row");
None
}
};
// v1.1.7 fix: fan the dead-letter out to matching handler triggers.
// This was missing since v1.1.1 — the row was written but
// `list_matching_dead_letter` had no production caller, so
// registered dead_letter handlers never fired. The recursion-stop
// (a dead-letter handler's own failure is not re-dead-lettered)
// is upheld by the `is_dead_letter_handler` short-circuit at the
// top of this function, so this fan-out is only reached for
// non-handler executions.
if let Some(dl_id) = dl_id {
self.fan_out_dead_letter(&row, &resolved, dl_id, &source, attempt, &err, now)
.await;
}
self.outbox
.delete(row.id)
.await
@@ -489,6 +507,82 @@ impl Dispatcher {
Ok(())
}
/// Enqueue one outbox row per matching `dead_letter` trigger so its
/// handler script runs with the dead-letter event as `ctx.event`.
/// Best-effort: a lookup/insert failure is logged, not propagated
/// (the dead-letter row itself is already durably written).
#[allow(clippy::too_many_arguments)]
async fn fan_out_dead_letter(
&self,
row: &OutboxRow,
resolved: &ResolvedTrigger,
dead_letter_id: DeadLetterId,
source: &str,
attempt: u32,
err: &ExecError,
now: DateTime<Utc>,
) {
// The DL event nests the original verbatim; if the payload can't
// be decoded back into a TriggerEvent we can't build the nested
// `original`, so skip the fan-out (the DL row is still written).
let Ok(original) = serde_json::from_value::<TriggerEvent>(row.payload.clone()) else {
tracing::warn!(
outbox_id = %row.id,
"dead-letter payload is not a TriggerEvent; skipping handler fan-out"
);
return;
};
let matches = match self
.triggers
.list_matching_dead_letter(row.app_id, source, row.trigger_id, Some(resolved.script_id))
.await
{
Ok(m) => m,
Err(e) => {
tracing::error!(?e, "dead-letter trigger lookup failed");
return;
}
};
for m in matches {
let event = TriggerEvent::DeadLetter {
dead_letter_id,
original: Box::new(original.clone()),
attempts: attempt,
last_error: err.to_string(),
trigger_id: row.trigger_id,
script_id: Some(resolved.script_id),
first_attempt_at: row.created_at,
last_attempt_at: now,
};
let payload = match serde_json::to_value(&event) {
Ok(p) => p,
Err(e) => {
tracing::error!(?e, "failed to serialize dead-letter event");
continue;
}
};
if let Err(e) = self
.outbox
.insert(NewOutboxRow {
app_id: row.app_id,
source_kind: OutboxSourceKind::DeadLetter,
trigger_id: Some(m.trigger_id),
script_id: Some(m.script_id),
reply_to: None,
payload,
origin_principal: Some(m.registered_by_principal),
trigger_depth: row.trigger_depth.saturating_add(1),
root_execution_id: row.root_execution_id,
})
.await
{
tracing::error!(?e, "failed to enqueue dead-letter handler delivery");
}
}
}
async fn deliver_inbox(&self, row: &OutboxRow, inbox_id: Uuid, result: InboxResult) {
match self.inbox.deliver(inbox_id, result.clone()).await {
InboxDeliveryOutcome::Delivered => {}