From fa6627ce7069983c62c2562ecb65e7de4f82be8f Mon Sep 17 00:00:00 2001
From: Ilya Kreymer <ikreymer@users.noreply.github.com>
Date: Wed, 12 Jun 2024 10:51:35 -0700
Subject: [PATCH] ensure QA configmap is updated for long running QA runs:
 (#1865)

- add a 'expire_at_duration_seconds' which is 75% of actual presign
duration time, or <25% remaining until presigned URL actually expires to
ensure presigned URLs are updated early than when they actually expire
- set cached expireAt time to the renew at time for more frequent
updates
- update QA configmap in place with updated presigned URLs when expireAt
time is reached
- mount qa config volume under /tmp/qa/ without subPath to get automatic
updates, which crawler will handle
- tests: fix qa test typo (from main)
- fixes #1864
---
 backend/btrixcloud/basecrawls.py      |  8 ++++++--
 backend/btrixcloud/operator/crawls.py | 27 +++++++++++++++++++++++++--
 backend/test/test_qa.py               |  2 +-
 chart/app-templates/crawler.yaml      |  5 ++---
 chart/templates/operators.yaml        |  2 +-
 5 files changed, 35 insertions(+), 9 deletions(-)

diff --git a/backend/btrixcloud/basecrawls.py b/backend/btrixcloud/basecrawls.py
index cf03bdea..5aec6be4 100644
--- a/backend/btrixcloud/basecrawls.py
+++ b/backend/btrixcloud/basecrawls.py
@@ -64,7 +64,8 @@ class BaseCrawlOps:
     background_job_ops: BackgroundJobOps
     page_ops: PageOps
 
-    presign_duration: int
+    presign_duration_seconds: int
+    expire_at_duration_seconds: int
 
     def __init__(
         self,
@@ -95,6 +96,9 @@ class BaseCrawlOps:
             min(presign_duration_minutes, PRESIGN_MINUTES_MAX) * 60
         )
 
+        # renew when <25% of time remaining
+        self.expire_at_duration_seconds = int(self.presign_duration_seconds * 0.75)
+
     def set_page_ops(self, page_ops):
         """set page ops reference"""
         self.page_ops = page_ops
@@ -434,7 +438,7 @@ class BaseCrawlOps:
             print("no files")
             return []
 
-        delta = timedelta(seconds=self.presign_duration_seconds)
+        delta = timedelta(seconds=self.expire_at_duration_seconds)
 
         out_files = []
 
diff --git a/backend/btrixcloud/operator/crawls.py b/backend/btrixcloud/operator/crawls.py
index 9962c827..e6eee4df 100644
--- a/backend/btrixcloud/operator/crawls.py
+++ b/backend/btrixcloud/operator/crawls.py
@@ -325,8 +325,15 @@ class CrawlOperator(BaseOperator):
         qa_source_crawl_id = params["qa_source_crawl_id"]
         name = f"qa-replay-{qa_source_crawl_id}"
 
-        if name in children[CMAP]:
-            return [children[CMAP][name]]
+        configmap = children[CMAP].get(name)
+        if configmap and not self._qa_configmap_update_needed(name, configmap):
+            metadata = configmap["metadata"]
+            configmap["metadata"] = {
+                "name": metadata["name"],
+                "namespace": metadata["namespace"],
+                "labels": metadata["labels"],
+            }
+            return [configmap]
 
         crawl_replay = await self.crawl_ops.get_internal_crawl_out(qa_source_crawl_id)
 
@@ -364,6 +371,22 @@ class CrawlOperator(BaseOperator):
 
         return self.load_from_yaml("crawler.yaml", params)
 
+    def _qa_configmap_update_needed(self, name, configmap):
+        try:
+            now = dt_now()
+            resources = json.loads(configmap["data"]["qa-config.json"])["resources"]
+            for resource in resources:
+                expire_at = datetime.fromisoformat(resource["expireAt"])
+                if expire_at <= now:
+                    print(f"Refreshing QA configmap for QA run: {name}")
+                    return True
+
+        # pylint: disable=broad-exception-caught
+        except Exception as e:
+            print(e)
+
+        return False
+
     # pylint: disable=too-many-arguments
     async def _resolve_scale(
         self,
diff --git a/backend/test/test_qa.py b/backend/test/test_qa.py
index 85f09e1d..e9861b31 100644
--- a/backend/test/test_qa.py
+++ b/backend/test/test_qa.py
@@ -383,7 +383,7 @@ def test_run_qa_not_running(
     count = 0
     while count < MAX_ATTEMPTS:
         r = requests.get(
-            f"{API_PREFIX}/orgs/{default_org_id}/crawls/{crawler_crawl_id}/activeQA",
+            f"{API_PREFIX}/orgs/{default_org_id}/crawls/{crawler_crawl_id}/qa/activeQA",
             headers=crawler_auth_headers,
         )
         data = r.json()
diff --git a/chart/app-templates/crawler.yaml b/chart/app-templates/crawler.yaml
index acd6e1b9..6dc618d0 100644
--- a/chart/app-templates/crawler.yaml
+++ b/chart/app-templates/crawler.yaml
@@ -124,7 +124,7 @@ spec:
         - {{ redis_url }}
       {% if qa_source_crawl_id %}
         - --qaSource
-        - /tmp/qa-config.json
+        - /tmp/qa/qa-config.json
       {% elif profile_filename %}
         - --profile
         - "@{{ profile_filename }}"
@@ -137,8 +137,7 @@ spec:
 
       {% if qa_source_crawl_id %}
         - name: qa-config
-          mountPath: /tmp/qa-config.json
-          subPath: qa-config.json
+          mountPath: /tmp/qa/
           readOnly: True
       {% endif %}
 
diff --git a/chart/templates/operators.yaml b/chart/templates/operators.yaml
index b7126edb..92e6a3a8 100644
--- a/chart/templates/operators.yaml
+++ b/chart/templates/operators.yaml
@@ -23,7 +23,7 @@ spec:
   - apiVersion: v1
     resource: configmaps
     updateStrategy:
-      method: OnDelete
+      method: InPlace
 
   hooks:
     sync: