Forum

i guess you meet similar issue with this post. here is my solution and i hope it can help you.

related post:

https://zeabur.com/forum/posts/69d50cf6ddd5e9c263d318b7

Environment:

• Provider: Alibaba Cloud (Aliyun) VPS

• Specs: 4 vCPUs / 8GB RAM

Issue: The system experienced CPU overloading during the initial startup phase, preventing OpenClaw from initializing correctly when other services were running simultaneously.

Solution: I temporarily stopped other resource-intensive applications to prioritize OpenClaw's bootstrapping. Once OpenClaw was fully operational, I manually restarted the remaining services. This sequential startup approach successfully bypassed the CPU spike and resolved the issue. It appears that 4 vCPUs may hit a bottleneck during concurrent service initialization.

Hi T Wen,

We found the crash cause in your runtime logs:

Config invalid
File: ~/.openclaw/openclaw.json
Problem:
 - hooks: Unrecognized key: "allowedAgentIds"

Your openclaw.json config has a key (allowedAgentIds inside the hooks section) that your current OpenClaw version (2026.2.2) doesn't recognize. OpenClaw validates its config on startup and exits immediately if it finds any unrecognized keys — that's what's causing the crash loop.

To fix, pick one of these:

1. Update your OpenClaw image to a newer version (Settings → Source → Image → change 2026.2.2 to the latest). Newer versions likely support the allowedAgentIds key, so the config would become valid without any edits.

2. Or remove the key from your config manually: use the Zeabur file browser on the service (or SSH into your server), open ~/.openclaw/openclaw.json, find the "allowedAgentIds" entry inside the "hooks" section (around line 301), and delete it. Save the file, then resume the service.

After either fix, resume your service from the dashboard and it should start up cleanly. Your server has plenty of free resources (2.4 GB RAM available, low load), so it's not a resource issue.