Store image data in pandas from dry runs

for easier poking around what will/should be culled

checkpoint

checkpoint

checkpoint

checkpoint

Author: minrk

scripts/delete-old-images.py

@@ -27,6 +27,7 @@
 
 from aiohttp_client_cache import CachedSession, SQLiteBackend
 import aiohttp
+import pandas as pd
 import tqdm
 import tqdm.asyncio
 import yaml
@@ -150,6 +151,7 @@ async def delete_image(session, image, digest, tags, dry_run=False):
     if dry_run:
         fetch = session.get
         verb = "Checking"
+        return
     else:
         fetch = session.delete
         verb = "Deleting"
@@ -246,7 +248,7 @@ async def bounded(f, *args, **kwargs):
 
     async with CachedSession(
         connector=aiohttp.TCPConnector(limit=2 * concurrency),
-        cache=SQLiteBackend(expire_after=24 * 3600),
+        cache=SQLiteBackend(expire_after=72 * 3600),
         **auth_kwargs,
     ) as session:
 
@@ -262,6 +264,8 @@ def should_delete_repository(image):
                 ci_string in image for ci_string in CI_STRINGS
             ):
                 return False
+            else:
+                return True
 
         def should_fetch_repository(image):
             if not any(substring in image for substring in R2D_STRINGS):
@@ -294,11 +298,11 @@ def should_fetch_repository(image):
             raise RuntimeError(
                 f"No images matching prefix {prefix}. Would delete all images!"
             )
-        print(f"Not deleting {repos_to_keep} images starting with {prefix}")
+        print(f"Not deleting {repos_to_keep} repos starting with {prefix}")
         if not tag_futures:
             print("Nothing to delete")
             return
-        print(f"{len(tag_futures)} images to delete (not counting tags)")
+        print(f"{len(tag_futures)} repos to consider for deletion (not counting tags)")
 
         delete_futures = set()
         done = set()
@@ -325,41 +329,64 @@ def should_delete_tag(image, info):
                 return False
 
             # check cutoff
-            image_ms = int(info["timeCreatedMs"])
+            image_ms = int(info["timeUploadedMs"])
             image_datetime = datetime.fromtimestamp(image_ms / 1e3)
             # sanity check timestamps
             if image_datetime < FIVE_YEARS_AGO or image_datetime > TOMORROW:
                 raise RuntimeError(
                     f"Not deleting image with weird date: {image}, {info}, {image_datetime}"
                 )
             if delete_before_ms > image_ms:
-                # if dry_run:
-                #     print(
-                #         f"\nWould delete {image}:{','.join(info['tag'])} {image_datetime.isoformat()}"
-                #     )
                 return True
             else:
                 return False
 
+        def save_stats():
+            df = pd.DataFrame(rows, columns=[
+                "image",
+                "digest",
+                "tags",
+                "size",
+                "date",
+            ])
+            today = datetime.today()
+            fname = f"registry-{release}-{today.strftime('%Y-%m-%d')}.pkl"
+            df.to_pickle(fname)
+
+        rows = []
         try:
             for f in tqdm.tqdm(
                 asyncio.as_completed(tag_futures),
                 total=len(tag_futures),
                 position=1,
-                desc="images retrieved",
+                desc="repos retrieved",
             ):
                 manifest = await f
                 image = manifest["name"]
                 delete_whole_repo = should_delete_repository(image)
                 if delete_whole_repo and len(manifest["manifest"]) > 1:
                     delete_progress.total += len(manifest["manifest"]) - 1
+
                 for digest, info in manifest["manifest"].items():
+                    image_ms = int(info["timeUploadedMs"])
+                    image_datetime = datetime.fromtimestamp(image_ms / 1e3)
+                    nbytes = int(info["imageSizeBytes"])
+                    rows.append(
+                        (
+                            image,
+                            digest,
+                            ",".join(info["tag"]),
+                            nbytes,
+                            image_datetime,
+                        )
+                    )
+                    if len(rows) % 100 == 0:
+                        save_stats()
                     if not should_delete_tag(image, info):
                         continue
                     if not delete_whole_repo:
                         # not counted yet
                         delete_progress.total += 1
-                    nbytes = int(info["imageSizeBytes"])
                     delete_byte_progress.total += nbytes
                     f = asyncio.ensure_future(
                         bounded(
@@ -389,6 +416,8 @@ def should_delete_tag(image, info):
             if delete_futures:
                 await asyncio.gather(*delete_futures)
         finally:
+            save_stats()
+
             delete_progress.close()
             delete_byte_progress.close()
             print("\n\n\n\n")