| name | troubleshoot-nvidia-gpu |
| description | Use when diagnosing issues with Nvidia Gpu: memory exhaustion (oom), thermal runaway then throttling cascade, silent hardware degradation, interconnect degradation (straggler), or gpu hang / fallen off bus. Queries Netdata via MCP for gpu reachability (driver communication), management path latency / driver health, gpu die temperature, hbm memory temperature, framebuffer memory utilization (vram), applies the diagnostic tree from the Netdata operator playbook, and recommends remediation. |
| version | 0.1.0 |
| author | Netdata |
| license | Apache-2.0 |
| tags | ["netdata","troubleshoot","mcp","nvidia-gpu"] |
Troubleshoot Nvidia Gpu
When to use this skill
- Memory exhaustion (OOM): Allocation fails immediately. No gradual degradation; it's a cliff.
The most common GPU workload failure.
- Thermal runaway then throttling cascade: Temperature rises then clocks throttle then
performance collapses then system finds equilibrium
at terrible performance. Workload continues but 2-10x
slower.
- Silent hardware degradation: ECC errors accumulate, memory corrupts intermittently, training
produces wrong results before anyone notices. The GPU "works" but
produces garbage.
- Interconnect degradation (straggler): One slow NVLink or degraded PCIe link in a distributed
training job creates a straggler that bottleneck-gates
the entire collective operation.
- GPU hang / fallen off bus: GPU becomes unresponsive, nvidia-smi hangs, requires driver reload
or reboot. XID 79 = catastrophic PCIe link failure.
- Driver/process hang: Stuck CUDA contexts hold GPU memory indefinitely. Zombie processes
prevent resource reclamation. May require
nvidia-smi -r or reboot.
- Any time the user reports a Nvidia Gpu service behaving outside its expected envelope (elevated
errors, latency, saturation, resource exhaustion, or unexpected restarts).
- An on-call engineer is paging on a Netdata alert tied to a Nvidia Gpu instance and wants a
structured triage path.
Key facts
- This skill wraps the Netdata operator playbook for Nvidia Gpu. It does not replace the playbook;
it routes a coding agent through MCP queries against the same signals the playbook relies on.
- The playbook decomposes Nvidia Gpu health into 8 signal domains: Availability, Thermal & Power,
Memory, Compute & Utilization, Interconnect, Errors (Xid). Each domain maps to one rule file in
this skill.
- Dominant failure archetypes the playbook calls out: Memory exhaustion (OOM); Thermal runaway then
throttling cascade; Silent hardware degradation; Interconnect degradation (straggler); GPU hang /
fallen off bus.
- Netdata observes the signals listed in the rule files via its native collectors, plus any
OpenTelemetry-shipped metrics that your Nvidia Gpu instrumentation adds. Both paths end at the
same MCP query surface.
- Netdata's nvidia_smi collector emits 18 context(s) under
nvidia_smi.*. The rule files enumerate
which contexts surface which domain; the Verification section below names the load-bearing ones
explicitly.
Step-by-step
- Confirm the Nvidia Gpu service is up. Query Netdata via MCP with
list_nodes and filter by the
host running the target. A missing node means the symptom is at the network or orchestrator
layer, not inside the service.
- Pull the last 15 minutes of signals for the target. Use
query_metrics against the contexts
listed in the domain rule files. Run find_anomalous_metrics in parallel over the same window;
anomalies frame which rule file to read first.
- Check for Memory exhaustion (OOM). Allocation fails immediately. No gradual degradation; it's
a cliff. The most common GPU workload failure. Inspect the rule file whose signals move first for
this mode.
- Check for Thermal runaway then throttling cascade. Temperature rises then clocks throttle
then performance collapses then system finds equilibrium at terrible performance. Workload
continues but 2-10x slower. Inspect the rule file whose signals move first for this mode.
- Check for Silent hardware degradation. ECC errors accumulate, memory corrupts intermittently,
training produces wrong results before anyone notices. The GPU "works" but produces garbage.
Inspect the rule file whose signals move first for this mode.
- Check for Interconnect degradation (straggler). One slow NVLink or degraded PCIe link in a
distributed training job creates a straggler that bottleneck-gates the entire collective
operation. Inspect the rule file whose signals move first for this mode.
- Check for GPU hang / fallen off bus. GPU becomes unresponsive, nvidia-smi hangs, requires
driver reload or reboot. XID 79 = catastrophic PCIe link failure. Inspect the rule file whose
signals move first for this mode.
- Correlate with host-level signals (
system.cpu.utilization, system.memory.usage,
system.disk.io_time). Many service-level failures have a host-resource precursor.
- Apply the remediation hinted at in the matching rule file or the operator playbook. Re-run the
MCP queries from the Verification section to confirm the signals returned to expected ranges. A
fix that does not move the signal back is not a fix.
Handy MCP call templates
# Discover metrics from Nvidia Gpu
list_metrics with q="nvidia_smi"
# Pull a specific context over the last window
query_metrics with context="nvidia_smi.gpu_mig_mode_current_status", relative_window=-15m
# Rank anomalies for the service or host
find_anomalous_metrics with node=<host> and context_pattern="nvidia_smi.*"
# Correlate a known problem context with others
find_correlated_metrics around the incident window
# Show current alert state
list_raised_alerts scoped to the node
Common mistakes
- Treating Nvidia Gpu as a generic HTTP or process health check. Nvidia Gpu has specific failure
archetypes (see Key facts) that generic checks miss.
- Stopping at the first anomalous metric. Several archetypes produce correlated spikes; use
find_correlated_metrics to widen the search before concluding a root cause.
- Quoting percentile latency without the sample count. Low traffic plus a single slow request moves
p99 by seconds.
- Reading dashboards for a window shorter than the failure's fingerprint. Slow-brew failures (queue
growth, bloat, memory fragmentation) need 30+ minutes of data to see the trend.
- Skipping the host-level correlation. A process-level fix for a noisy-neighbour problem does not
hold.
- Assuming alert thresholds are tuned for your workload. Tune against observed Nvidia Gpu traffic
before escalating an alert configuration issue.
Verification
Run these MCP queries against the Netdata instance that sees the Nvidia Gpu service. Every context
listed below is a real Netdata chart name; the agent does not need to guess.
1. list_metrics filtered by q="nvidia_smi" (returns every nvidia_smi.* context Netdata sees)
2. query_metrics with contexts=[nvidia_smi.gpu_mig_mode_current_status, nvidia_smi.gpu_pcie_bandwidth_usage, nvidia_smi.gpu_memory_utilization, nvidia_smi.gpu_frame_buffer_memory_usage, nvidia_smi.gpu_bar1_memory_usage, nvidia_smi.gpu_mig_frame_buffer_memory_usage] and relative_window=-30m
3. find_anomalous_metrics filtered by node=<host> and context_pattern="nvidia_smi.*"
Load-bearing contexts for this service:
nvidia_smi.gpu_mig_mode_current_status: MIG current mode (status). Dimensions: enabled,
disabled.
nvidia_smi.gpu_pcie_bandwidth_usage: PCI Express Bandwidth Usage (B/s). Dimensions: rx, tx.
nvidia_smi.gpu_memory_utilization: Memory utilization (%). Dimensions: memory.
nvidia_smi.gpu_frame_buffer_memory_usage: Frame buffer memory usage (B). Dimensions: free, used,
reserved.
nvidia_smi.gpu_bar1_memory_usage: BAR1 memory usage (B). Dimensions: free, used.
nvidia_smi.gpu_mig_frame_buffer_memory_usage: Frame buffer memory usage (B). Dimensions: free,
used, reserved.
A clean result means every context is within its expected band and the find_anomalous_metrics list
is empty or contains only already-acknowledged items. If the fix was real, re-running the same
queries 10 minutes after applying it will show a clean result. If it does not, revert and look
deeper.
When the fix does not hold
If signals drift back into the anomalous range within 30 minutes of a remediation, the cause was
deeper than the applied change. Typical misdiagnoses for Nvidia Gpu:
- Host-resource pressure masquerading as application bug.
- Dependent service (DB, cache, upstream) causing a secondary symptom in the instrumented service.
- Configuration change that was never reloaded (some subsystems only pick up config on full
restart).
Escalate by widening the query window: 2-6 hours instead of 15 minutes. Slow-moving causes are
invisible at triage window sizes.
References