Skip to content

SIEM troubleshooting

The Guardian errors you are most likely to meet when shipping logs, searching, and running rules, triggers and sends, with the cause and the fix.

Updated 04 Nov 2026Markdown

Guardian’s console API answers errors as { "error": "<message>", "code": "<code>" }; the Resolve Platform shows the message. Ingest answers senders with a status code that Firehose and Logpush act on, and records the reason as the token’s Last error on Guardian > Sources. The codes are listed on API.

Access and plan

Symptom or message Cause Fix
Guardian is an add-on to Business. Choose an ingest pack to start. (402, needs: pack), or Guardian needs the Business plan. (needs: business) No active ingest pack, or the plan is below Business An owner or admin presses Start Guardian. See Packs and limits.
Sign in again. (401, code sign_in) The Resolve Platform session’s access token is missing, invalid or expired Sign in to the Resolve Platform again. Vulnesis API keys are not accepted on the Guardian API.

Shipping logs

Symptom or message Cause Fix
401 at ingest (Firehose parks batches; Logpush marks the job failing) The token is unknown, revoked, or of another kind. An aws token on the Cloudflare or events route is refused. Create a token of the matching kind and update the stack parameter or the job destination. See Kinds.
Last error: Today's ingest is over one and a half times the pack's daily cap. ... (429, ingest_cap) Today’s ingest passed 150% of the pack’s cap Wait until 00:00 UTC, choose a larger pack, or drop a high-volume dataset such as http_requests. See The daily ingest cap.
Last error: <n> flow records were not in a format Guardian reads The flow log uses a custom format Guardian does not parse Use the stack’s flow log format or the default 14-field format. See Flow log format.
cloud_audit is empty or has no read calls EventBridge gets CloudTrail events only while a trail logs them, and never read-only calls Make sure a trail logs management events in the region, and deploy the stack with SendCloudTrail=true in every region. See CloudTrail needs a trail.
Cloudflare: The zone's <plan> plan does not include Logpush for <dataset>. ... Cloudflare refused the job, often worded as “exceeded max jobs allowed” Push only datasets your Cloudflare plan offers. See Datasets.

Searching

Symptom or message Cause Fix
field: "<name>" is not a column here; available: ... A misspelt column, or a column dropped by an earlier stats or project Use a listed column, or read a non-column field with raw.<key>. See Search syntax.
Today's query budget is used up. It resets at 00:00 UTC. (429, budget) Searches, rules, triggers and sends have scanned the pack’s daily budget Wait for 00:00 UTC, narrow ranges, or choose a larger pack. Usage shows what scanned the most. See The query budget.

Rules, triggers and sends

Symptom or message Cause Fix
A rule is invalid, or the save fails with code compile The rule no longer compiles, or uses something Guardian refuses: a field the table does not store, an unsupported aggregation, a chained correlation Read the reason beside the rule. See What is refused.
the plan runs at most <n> rules at once (code limit; also triggers, and the plan keeps at most <n> scheduled sends) The pack’s cap on enabled rules, triggers or sends is reached Turn one off, or choose a larger pack. See Packs.
a trigger still uses this rule; remove it first, or a scheduled send or a trigger still uses this search; remove those first (code in_use) The rule or search is a trigger’s or send’s source Delete or repoint the trigger or send first
A rule goes erroring, then disabled_error; siem.rule.failed arrives The warehouse refused the rule’s query three, then eight, times in a row Open Runs for the error, fix the rule, and turn it on again. See Statuses.
A trigger never fires No group reached minRows in windowSec, or it is in cooldownSec, outside hours, or past maxPerHour Run Dry run to see how often the levers would have fired in the last 24 hours. See Dry run.
A scheduled send never arrives whenEmpty is skip and the search returned no rows Set whenEmpty to all_clear, or Preview the next message. See Rows.

Still stuck

Collect the exact error message and code, the request id Firehose recorded for a failed batch, the token prefix and Last error from Guardian > Sources, and for a rule its last entries from Runs (GET /v1/guardian/rules/{id}/runs). For a search, Show SQL gives the query Guardian ran.

esc
↑ ↓ move↵ openSearches in your browser. A key like autonomy.max or a code like threshold_required jumps straight to it.