Deployment & Kubernetes

Pod, image and cluster issues when running h2oGPTe on Kubernetes.

Fix: h2oGPTe agent pod restarted by Kubernetes liveness/readiness probes under load
At a glance Problem Agent pod is killed and restarted under concurrent-session load; separately, the /budget endpoint fails. Affects Enterprise h2oGPTe 1...
Wed, 5 Aug, 2026 at 6:09 AM
Fix: vLLM pod crashes at NCCL init with 'operation not supported' on NVIDIA vGPU
At a glance Problem The vLLM container crashes on startup during NCCL initialisation and never serves a request. Occurs only when the model is spread acros...
Tue, 1 Sep, 2026 at 2:04 AM