allow disabling of auto-resize of crawler pods (#1964)

- only enable if 'enable_auto_resize' is true, default to false
- if true, set memory limit to 1.2 of memory requests, resize when
hitting 'soft oom' of initial request, adjust by 1.2 (current behavior)
up to max_crawler_memory
- if false, set memory limit to max_crawler_memory and never adjust
memory requests or memory limits
- part of #1959

backend/btrixcloud/operator/baseoperator.py

@@ -7,6 +7,7 @@ from kubernetes.utils import parse_quantity
 
 import yaml
 from btrixcloud.k8sapi import K8sAPI
+from btrixcloud.utils import is_bool
 
 
 if TYPE_CHECKING:
@@ -30,6 +31,7 @@ class K8sOpAPI(K8sAPI):
     """Additional k8s api for operators"""
 
     has_pod_metrics: bool
+    enable_auto_resize: bool
     max_crawler_memory_size: int
 
     def __init__(self):
@@ -39,6 +41,7 @@ class K8sOpAPI(K8sAPI):
             self.shared_params = yaml.safe_load(fh_config)
 
         self.has_pod_metrics = False
+        self.enable_auto_resize = False
         self.max_crawler_memory_size = 0
 
         self.compute_crawler_resources()
@@ -127,6 +130,11 @@ class K8sOpAPI(K8sAPI):
         self.has_pod_metrics = await self.is_pod_metrics_available()
         print("Pod Metrics Available:", self.has_pod_metrics)
 
+        self.enable_auto_resize = self.has_pod_metrics and is_bool(
+            os.environ.get("ENABLE_AUTO_RESIZE_CRAWLERS")
+        )
+        print("Auto-Resize Enabled", self.enable_auto_resize)
+
 
 # pylint: disable=too-many-instance-attributes, too-many-arguments
 # ============================================================================

backend/btrixcloud/operator/crawls.py

@@ -29,11 +29,7 @@ from btrixcloud.models import (
     StorageRef,
 )
 
-from btrixcloud.utils import (
-    from_k8s_date,
-    to_k8s_date,
-    dt_now,
-)
+from btrixcloud.utils import from_k8s_date, to_k8s_date, dt_now
 
 from .baseoperator import BaseOperator, Redis
 from .models import (
@@ -84,6 +80,13 @@ MEM_LIMIT_PADDING = 1.2
 class CrawlOperator(BaseOperator):
     """CrawlOperator Handler"""
 
+    done_key: str
+    pages_key: str
+    errors_key: str
+
+    fast_retry_secs: int
+    log_failed_crawl_lines: int
+
     def __init__(self, *args):
         super().__init__(*args)
 
@@ -221,8 +224,7 @@ class CrawlOperator(BaseOperator):
                 data.related.get(METRICS, {}),
             )
 
-            # auto-scaling not possible without pod metrics
-            if self.k8s.has_pod_metrics:
+            if self.k8s.enable_auto_resize:
                 # auto sizing handled here
                 await self.handle_auto_size(status.podStatus)
 
@@ -377,7 +379,10 @@ class CrawlOperator(BaseOperator):
         params["priorityClassName"] = pri_class
         params["cpu"] = pod_info.newCpu or params.get(cpu_field)
         params["memory"] = pod_info.newMemory or params.get(mem_field)
-        params["memory_limit"] = float(params["memory"]) * MEM_LIMIT_PADDING
+        if self.k8s.enable_auto_resize:
+            params["memory_limit"] = float(params["memory"]) * MEM_LIMIT_PADDING
+        else:
+            params["memory_limit"] = self.k8s.max_crawler_memory_size
         params["workers"] = params.get(worker_field) or 1
         params["do_restart"] = (
             pod_info.should_restart_pod() or params.get("force_restart")
@@ -555,7 +560,7 @@ class CrawlOperator(BaseOperator):
             },
         ]
 
-        if self.k8s.has_pod_metrics:
+        if self.k8s.enable_auto_resize:
             related_resources.append(
                 {
                     "apiVersion": METRICS_API,
@@ -1072,7 +1077,7 @@ class CrawlOperator(BaseOperator):
             pod_info.used.storage = storage
 
             # if no pod metrics, get memory estimate from redis itself
-            if not self.k8s.has_pod_metrics:
+            if not self.k8s.enable_auto_resize:
                 stats = await redis.info("memory")
                 pod_info.used.memory = int(stats.get("used_memory_rss", 0))
 

chart/templates/configmap.yaml

@@ -58,6 +58,8 @@ data:
 
   MAX_CRAWLER_MEMORY: "{{ .Values.max_crawler_memory }}"
 
+  ENABLE_AUTO_RESIZE_CRAWLERS: "{{ .Values.enable_auto_resize_crawlers }}"
+
   BILLING_ENABLED: "{{ .Values.billing_enabled }}"
 
   SALES_EMAIL: "{{ .Values.sales_email }}"

chart/values.yaml

@@ -262,6 +262,12 @@ crawler_extra_memory_per_browser: 768Mi
 # crawler_memory = crawler_memory_base + crawler_memory_per_extra_browser * (crawler_browser_instances - 1)
 # crawler_memory:
 
+# Crawler Autoscaling
+# ---------------------
+
+# if set to true, automatically adjust crawler memory usage up to max_crawler_memory
+enable_auto_resize_crawlers: false
+
 
 # max crawler memory, if set, will enable auto-resizing of crawler pods up to this size
 # if not set, no auto-resizing is done, and crawls always use 'crawler_memory' memory