Skip to content
Help: OpenClaw runt...
 
Notifications
Clear all

Help: OpenClaw runtime is eating our Azure budget. Anyone set up hard cost limits?

1 Posts
1 Users
0 Reactions
90 Views
(@frankd)
Reputable Member
Joined: 4 months ago
Posts: 313
Topic starter   [#18830]

Hello everyone. I've been a big proponent of OpenClaw's agentic response framework for our AI SOC, and the technical capabilities for automated investigation are genuinely impressive. However, we've hit a major operational snag that I suspect others might be facing, or hopefully, have already solved.

Our production deployment in Azure, handling medium-volume alert enrichment and triage, has begun generating unpredictable and frankly alarming compute costs. The runtime seems to spin up more container instances than we anticipate during peak alert batches, and the LLM calls for narrative generation, while powerful, are a separate cost vector that's difficult to forecast. Our last month's bill was nearly double what we'd projected, which is unsustainable.

I'm looking for practical strategies to enforce hard cost boundaries without completely crippling the system's functionality. We need it to throttle or queue, not just break catastrophically when limits are hit. Specifically:

* **Azure-specific budget tools:** Has anyone successfully used Azure Budgets with action groups to trigger automation (like a Logic App or Function) to scale down the OpenClaw deployment? Did you find a reliable metric or threshold?
* **Native OpenClaw configuration:** Are there any levers within OpenClaw's own orchestration engine or agent configuration to limit concurrent agent execution or set a hard cap on LLM token usage per investigation cycle?
* **Queue-and-throttle workflows:** I'm considering building an external dispatcher that holds alerts in a queue and meters them out to OpenClaw based on a monthly compute budget. This seems clunky. Has anyone implemented a similar pattern, perhaps with Azure Service Bus or Storage Queues?
* **Vendor comparison point:** For those who also evaluated commercial AI SOC platforms, did you find their pricing models (per alert, per user, flat SaaS) offered more predictable cost control, even if the upfront price was higher?

Our core requirement is to maintain the automated triage for common alert types without the risk of a runaway cost event. Any insights into your own cost control architectures, lessons learned from scaling, or even configuration snippets for Azure Policy would be immensely valuable to this community.

— frank


buyer beware, but buy smart


   
Quote