Skip to content

Commit 04d38f6

Browse files
committed
docs(serverless): document missing fitness-check env vars
- Add RUNPOD_GPU_MAX_ERROR_MESSAGES to advanced GPU check config - Document RUNPOD_SKIP_AUTO_SYSTEM_CHECKS and RUNPOD_SKIP_GPU_CHECK under a new "Disabling Built-in Checks" subsection, noting that user-registered checks still run when these flags are set These flags exist in rp_fitness.py and rp_gpu_fitness.py but were previously undocumented.
1 parent d1c3900 commit 04d38f6

1 file changed

Lines changed: 27 additions & 0 deletions

File tree

docs/serverless/worker_fitness_checks.md

Lines changed: 27 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -199,6 +199,12 @@ os.environ["RUNPOD_GPU_TEST_TIMEOUT"] = "60"
199199

200200
# Override binary path (for custom/patched versions)
201201
os.environ["RUNPOD_BINARY_GPU_TEST_PATH"] = "/custom/path/gpu_test"
202+
203+
# Cap the number of error messages parsed from gpu_test output (default: 10)
204+
os.environ["RUNPOD_GPU_MAX_ERROR_MESSAGES"] = "20"
205+
206+
# Skip auto-registration of this check (primarily for testing)
207+
os.environ["RUNPOD_SKIP_GPU_CHECK"] = "true"
202208
```
203209

204210
**What it tests**:
@@ -374,6 +380,27 @@ os.environ["RUNPOD_MIN_MEMORY_GB"] = "8.0"
374380
os.environ["RUNPOD_MIN_DISK_PERCENT"] = "15.0"
375381
```
376382

383+
### Disabling Built-in Checks
384+
385+
For testing or specialized deployments, built-in checks can be disabled via environment variables. These are not recommended for production use.
386+
387+
| Env var | Effect |
388+
|---|---|
389+
| `RUNPOD_SKIP_AUTO_SYSTEM_CHECKS=true` | Skips auto-registration of memory, disk, network, CUDA version, CUDA init, and GPU benchmark checks |
390+
| `RUNPOD_SKIP_GPU_CHECK=true` | Skips auto-registration of the native GPU memory allocation test (`gpu_test` binary) |
391+
392+
```python
393+
import os
394+
395+
# Disable all auto-registered system checks (testing only)
396+
os.environ["RUNPOD_SKIP_AUTO_SYSTEM_CHECKS"] = "true"
397+
398+
# Disable the automatic GPU memory allocation test
399+
os.environ["RUNPOD_SKIP_GPU_CHECK"] = "true"
400+
```
401+
402+
User-registered checks via `@register_fitness_check` still run regardless of these flags.
403+
377404
## Behavior
378405

379406
### Execution Timing

0 commit comments

Comments
 (0)