Environment:
- OS: Windows 11
- Quip Node Manager version: v0.2.6-rc1
- Run mode: Docker (GPU/cuda profile)
- GPU: NVIDIA GeForce RTX 3060 Laptop GPU
- Driver: 616.92, CUDA 13.4
What happened:
After updating to v0.2.6-rc1 (which fixed the earlier CUDA startup hang — thank you!), the node now starts and mines successfully on GPU, but the Status Monitor repeatedly cycles between "RUNNING" and "DEGRADED" every ~30-60 seconds.
Key detail: the Infrastructure line itself shows:
Infrastructure: ok — all services up
But the overall status still flips to DEGRADED, with:
Chain: unknown — POST quip-validator:9944/ via quip-validator: exit code 2; --: line 3: set: pipefail : invalid option name
Participation: unknown — rpc unreachable
This suggests the chain/participation health-check probe script itself is failing due to a shell incompatibility (set -o pipefail being run somewhere that doesn't support it — likely sh/dash instead of bash), not an actual failure of the validator or chain sync.
Evidence the node is actually healthy: Node Logs during this same window show the validator actively importing blocks and the miner actively compiling CUDA kernels and submitting mining attempts — i.e., real functional activity, contradicting the DEGRADED label.
Steps to reproduce:
- Update to v0.2.6-rc1 on Windows via the install script.
- Enable GPU mining, start the node.
- Watch Status Monitor — it will cycle between RUNNING and DEGRADED repeatedly, always citing the same pipefail error on the Chain/Participation checks.
Expected behavior:
Status should reflect actual container/service health, not fail due to an internal shell script bug in the health-check probe itself.
Suspicion: possibly a regression introduced alongside the fix in MR !62 / v0.2.6-rc1, since this specific probe/error wasn't present in v0.2.5 (though v0.2.5 had the separate CUDA-hang bug instead).
Environment:
What happened:
After updating to v0.2.6-rc1 (which fixed the earlier CUDA startup hang — thank you!), the node now starts and mines successfully on GPU, but the Status Monitor repeatedly cycles between "RUNNING" and "DEGRADED" every ~30-60 seconds.
Key detail: the Infrastructure line itself shows:
Infrastructure: ok — all services up
But the overall status still flips to DEGRADED, with:
Chain: unknown — POST quip-validator:9944/ via quip-validator: exit code 2; --: line 3: set: pipefail : invalid option name
Participation: unknown — rpc unreachable
This suggests the chain/participation health-check probe script itself is failing due to a shell incompatibility (
set -o pipefailbeing run somewhere that doesn't support it — likelysh/dash instead ofbash), not an actual failure of the validator or chain sync.Evidence the node is actually healthy: Node Logs during this same window show the validator actively importing blocks and the miner actively compiling CUDA kernels and submitting mining attempts — i.e., real functional activity, contradicting the DEGRADED label.
Steps to reproduce:
Expected behavior:
Status should reflect actual container/service health, not fail due to an internal shell script bug in the health-check probe itself.
Suspicion: possibly a regression introduced alongside the fix in MR !62 / v0.2.6-rc1, since this specific probe/error wasn't present in v0.2.5 (though v0.2.5 had the separate CUDA-hang bug instead).