Skip to content
Notifications
Clear all

TIL: How to extract and interpret Claw's internal metrics (they're buried).

2 Posts
2 Users
0 Reactions
62 Views
(@cloud_security_sera)
Honorable Member
Joined: 5 months ago
Posts: 543
Topic starter   [#18194]

Claw's vendor dashboard shows you nothing useful. Their "health score" is marketing fluff.

The real metrics are in the collector's local telemetry. They're exposed but undocumented. You need to pull them via the local admin port, which they don't tell you to secure.

```
# On the collector host:
curl -s http://localhost:8065/debug/metrics | jq .
```

Key fields to grep for:
* `claw_collector_errors_total`
* `claw_collector_missed_events`
* `claw_api_ratelimit_exceeded`

Their support will say "the dashboard is the source of truth." It's not. Found our collector was silently dropping 5% of logs due to throttling at their ingest endpoint. Their dashboard showed "All Systems Operational."

Would not renew without contractual SLA based on these internal metrics.


Least privilege is not a suggestion.


   
Quote
(@aarons)
Reputable Member
Joined: 4 months ago
Posts: 342
 

Good find on the collector metrics. That's the only real source of truth for your data integrity.

You're absolutely right to push for contractual SLAs based on them. I'd add `claw_storage_queue_depth` to your watch list. I've seen that spike before the vendor's ingest buffers overflow, and you'll see the "missed_events" counter jump ten minutes later.

Just be aware that in their hosted/SaaS collector model, that admin port isn't accessible. You have to rely on their API, which conveniently lags by hours and aggregates away the per-customer spikes. It's a feature, not a bug.


Your cloud bill is 30% too high


   
ReplyQuote