fix: prune non-successful detail tables unconditionally when rollup watermark is stale
All checks were successful
deploy / deploy (push) Successful in 52s

Same pattern as the raw quotes fix (previous commit), now extended to all
detail tables: swap_demand_events, trade_decisions, execute_trade_commands,
trade_execution_results, quote_outcome_attributions. The existing
pruneNonSuccessfulQuoteLifecycleHistory function was dead code — written,
tested, exported, never called.

Without this, trade_decisions and swap_demand_events grew to 17GB and 16GB
respectively in 42 hours because the rollup-watermark gate blocked all
detail pruning. The disk went from 28% to 81% in two days and was hours
from crashing the cluster again.

Successful evidence (completed outcomes, attributed settlements) is always
preserved by the prune predicates. Only non-successful detail older than
the retention window is pruned.

Proof: runQuoteLifecycleRetentionMaintenance now calls pruneNonSuccessfulQuoteLifecycleHistory before the watermark gate. All 313 tests pass. The stale-rollup test was updated to expect unconditional detail pruning while still asserting the rollup itself remains bounded.

Assumptions: pruneNonSuccessfulQuoteLifecycleHistory already preserves rows with successful outcomes via payload field checks and correlated existence checks against trade_execution_results and quote_outcome_attributions. Autovacuum will reclaim dead tuples for reuse; one-time TRUNCATE was used as emergency cleanup for the existing backlog.

Still fake: fromBeginning:true replay on restart still creates rollup backlog. local-path PVC sizes are not enforced (platform issue). No disk monitoring (platform issue). Rollup analytics will be incomplete during periods when the watermark is stale — non-successful detail is pruned before rollup coverage.
This commit is contained in:
philipp 2026-06-24 12:40:29 +02:00
parent 5e6555d0ae
commit 691c98a62d
2 changed files with 23 additions and 15 deletions

View file

@ -2579,26 +2579,34 @@ export async function runQuoteLifecycleRetentionMaintenance(pool, {
windowEnd: rollupCutoff,
});
// Prune raw quotes unconditionally — they are a debug firehose that must be
// pruned even when the rollup watermark is stale. Without this, disk
// pressure builds up and blocks the only cleanup that can relieve it.
// Prune raw quotes and non-successful detail unconditionally — even when
// the rollup watermark is stale. Successful evidence is always preserved.
// Without this, disk pressure builds up and blocks the only cleanup that
// can relieve it, creating a deadlock that ends in node-wide eviction.
const rawPruneResult = await pruneRawNearIntentsQuoteHistory(pool, {
now,
retainRecentMs: policy.raw_retention_ms,
batchSize: Math.min(500_000, Math.floor(Number(policy.max_delete_rows_per_pass) || 0)),
maxBatches: 10,
});
const detailPruneResult = await pruneNonSuccessfulQuoteLifecycleHistory(pool, {
now,
retainRecentMs: modeDecision.detail_retention_ms || policy.normal_detail_retention_ms,
batchSize: Math.min(500_000, Math.floor(Number(policy.max_delete_rows_per_pass) || 0)),
maxBatches: 10,
});
const mergedTables = new Map();
for (const t of [...(detailPruneResult.tables || []), { table: 'raw_near_intents_quotes', cutoff: rawPruneResult.cutoff, deletedCount: rawPruneResult.deletedCount }]) {
const prev = mergedTables.get(t.table);
mergedTables.set(t.table, prev ? { ...prev, deletedCount: Number(prev.deletedCount || 0) + Number(t.deletedCount || 0) } : t);
}
pruneResult = {
ok: true,
mode: modeDecision.mode,
deletedCount: rawPruneResult.deletedCount,
deletedCount: rawPruneResult.deletedCount + detailPruneResult.deletedCount,
detail_cutoff: detailCutoff,
raw_cutoff: rawPruneResult.cutoff,
tables: [{
table: 'raw_near_intents_quotes',
cutoff: rawPruneResult.cutoff,
deletedCount: rawPruneResult.deletedCount,
}],
tables: [...mergedTables.values()],
};
const watermark = await loadQuoteLifecycleRollupWatermark(pool);

View file

@ -337,8 +337,8 @@ test('retention maintenance bounds historical rollup catch-up and fails pruning
if (sql.includes('quote_id IS NULL')) return { rows: [], rowCount: 0 };
if (sql.includes('INSERT INTO quote_lifecycle_rollup_watermarks')) return { rows: [], rowCount: 1 };
if (sql.includes('INSERT INTO quote_lifecycle_retention_runs')) return { rows: [], rowCount: 1 };
if (sql.includes('DELETE FROM') && !sql.includes('raw_near_intents_quotes')) throw new Error('detail prune must not run before watermark reaches cutoff');
if (sql.includes('DELETE FROM') && sql.includes('raw_near_intents_quotes')) return { rows: [], rowCount: 0 };
// Unconditional pruning (preserving successful evidence) runs even when the watermark is stale.
if (sql.includes('DELETE FROM')) return { rows: [], rowCount: 0 };
return { rows: [], rowCount: 0 };
},
};
@ -357,9 +357,9 @@ test('retention maintenance bounds historical rollup catch-up and fails pruning
'2026-06-05T00:00:00.000Z',
'2026-06-05T04:00:00.000Z',
]);
// Raw quotes are pruned unconditionally; detail tables stay blocked until the watermark catches up.
// Unconditional TTL pruning runs for all tables even when the watermark is stale.
// Successful evidence is preserved by the prune predicates.
const deleteQueries = queries.filter((query) => query.sql.includes('DELETE FROM'));
assert.equal(deleteQueries.length, 1);
assert.equal(deleteQueries[0].sql.includes('raw_near_intents_quotes'), true);
assert.equal(deleteQueries.some((query) => query.sql.includes('DELETE FROM') && !query.sql.includes('raw_near_intents_quotes')), false);
assert.ok(deleteQueries.length >= 1, 'expected at least one DELETE query');
assert.ok(deleteQueries.every((q) => q.sql.includes('raw_near_intents_quotes') || q.sql.includes('swap_demand_events') || q.sql.includes('trade_decisions') || q.sql.includes('execute_trade_commands') || q.sql.includes('trade_execution_results') || q.sql.includes('quote_outcome_attributions')), true);
});