set ewma factor to 0.01

Author: elfkuzco

backend/src/zimfarm_backend/common/schemas/orms.py

@@ -115,7 +115,7 @@ class TaskResourceUsageSchema(BaseModel):
 
 
 class TaskCPUUsageSchema(TaskResourceUsageSchema):
-    ewma: float | None = Field(default=None, alias="avg")
+    avg_usage: float | None = Field(default=None, alias="avg")
 
 
 class TaskStatsSchema(BaseModel):

worker/src/zimfarm_worker/task/worker.py

@@ -39,7 +39,7 @@
 from zimfarm_worker.task.zim import get_zim_info
 
 SLEEP_INTERVAL = 60  # nb of seconds to sleep before watching
-CPU_EWMA_ALPHA = 0.25  # EWMA smoothing factor for CPU percentage samples (0..1)
+CPU_EWMA_ALPHA = 0.01  # EWMA smoothing factor for CPU percentage samples (0..1)
 
 
 PENDING = "pending"
@@ -123,7 +123,7 @@ def __init__(
 
         self.max_memory_usage: int = 0  # maximum memory used by scraper
         self.max_disk_usage: int = 0  # maximum disk used by scraper
-        self.cpu_ewma: float = 0.0  # cpu exponential moving weighted average
+        self.avg_cpu_usage: float = 0.0  # cpu exponential moving weighted average
         self.max_cpu_usage: float = 0.0  # maximum cpu percentage used by scraper
 
         # register stop/^C
@@ -230,11 +230,12 @@ def _compute_scraper_cpu_stats(self, scraper_stats: dict[str, Any]) -> float:
             cpu_sample = (delta_cpu / float(delta_system)) * float(online_cpus) * 100.0
 
         # apply EWMA smoothing to reduce effect of short spikes
-        if self.cpu_ewma == 0.0:
-            self.cpu_ewma = cpu_sample
+        if self.avg_cpu_usage == 0.0:
+            self.avg_cpu_usage = cpu_sample
         else:
-            self.cpu_ewma = (
-                CPU_EWMA_ALPHA * cpu_sample + (1.0 - CPU_EWMA_ALPHA) * self.cpu_ewma
+            self.avg_cpu_usage = (
+                CPU_EWMA_ALPHA * cpu_sample
+                + (1.0 - CPU_EWMA_ALPHA) * self.avg_cpu_usage
             )
         return cpu_sample
 
@@ -271,7 +272,7 @@ def submit_scraper_progress(self):
             },
             "cpu": {
                 "max_usage": self.max_cpu_usage,
-                "ewma": round(self.cpu_ewma, 2),
+                "avg_usage": round(self.avg_cpu_usage, 2),
             },
             "disk": {
                 "max_usage": self.max_disk_usage,