Fix cause of checksum failures in public binary mirror (#32407)

Move the copying of the buildcache to a root job that runs after all the child
pipelines have finished, so that the operation can be coordinated across all
child pipelines to remove the possibility of race conditions during potentially
simlutandous copies. This lets us ensure the .spec.json.sig and .spack files
for any spec in the root mirror always come from the same child pipeline
mirror (though which pipeline is arbitrary).  It also allows us to avoid copying
of duplicates, which we now do.

2 files changed, 113 insertions, 45 deletions

diff --git a/lib/spack/spack/ci.py b/lib/spack/spack/ci.py
index b1c777bb8b..198e787dea 100644
--- a/lib/spack/spack/ci.py
+++ b/lib/spack/spack/ci.py
@@ -36,6 +36,7 @@ import spack.repo
 import spack.util.executable as exe
 import spack.util.gpg as gpg_util
 import spack.util.spack_yaml as syaml
+import spack.util.url as url_util
 import spack.util.web as web_util
 from spack.error import SpackError
 from spack.reporters.cdash import CDash
@@ -644,8 +645,6 @@ def generate_gitlab_ci_yaml(
     # Values: "spack_pull_request", "spack_protected_branch", or not set
     spack_pipeline_type = os.environ.get("SPACK_PIPELINE_TYPE", None)
 
-    spack_buildcache_copy = os.environ.get("SPACK_COPY_BUILDCACHE", None)
-
     if "mirrors" not in yaml_root or len(yaml_root["mirrors"].values()) < 1:
         tty.die("spack ci generate requires an env containing a mirror")
 
@@ -653,6 +652,12 @@ def generate_gitlab_ci_yaml(
     mirror_urls = [url for url in ci_mirrors.values()]
     remote_mirror_url = mirror_urls[0]
 
+    spack_buildcache_copy = os.environ.get("SPACK_COPY_BUILDCACHE", None)
+    if spack_buildcache_copy:
+        buildcache_copies = {}
+        buildcache_copy_src_prefix = remote_mirror_override or remote_mirror_url
+        buildcache_copy_dest_prefix = spack_buildcache_copy
+
     # Check for a list of "known broken" specs that we should not bother
     # trying to build.
     broken_specs_url = ""
@@ -1020,6 +1025,36 @@ def generate_gitlab_ci_yaml(
                         "{0} ({1})".format(release_spec, release_spec_dag_hash)
                     )
 
+                # Only keep track of these if we are copying rebuilt cache entries
+                if spack_buildcache_copy:
+                    # TODO: This assumes signed version of the spec
+                    buildcache_copies[release_spec_dag_hash] = [
+                        {
+                            "src": url_util.join(
+                                buildcache_copy_src_prefix,
+                                bindist.build_cache_relative_path(),
+                                bindist.tarball_name(release_spec, ".spec.json.sig"),
+                            ),
+                            "dest": url_util.join(
+                                buildcache_copy_dest_prefix,
+                                bindist.build_cache_relative_path(),
+                                bindist.tarball_name(release_spec, ".spec.json.sig"),
+                            ),
+                        },
+                        {
+                            "src": url_util.join(
+                                buildcache_copy_src_prefix,
+                                bindist.build_cache_relative_path(),
+                                bindist.tarball_path_name(release_spec, ".spack"),
+                            ),
+                            "dest": url_util.join(
+                                buildcache_copy_dest_prefix,
+                                bindist.build_cache_relative_path(),
+                                bindist.tarball_path_name(release_spec, ".spack"),
+                            ),
+                        },
+                    ]
+
                 if artifacts_root:
                     job_dependencies.append(
                         {"job": generate_job_name, "pipeline": "{0}".format(parent_pipeline_id)}
@@ -1197,32 +1232,6 @@ def generate_gitlab_ci_yaml(
 
             output_object["sign-pkgs"] = signing_job
 
-        if spack_buildcache_copy:
-            # Generate a job to copy the contents from wherever the builds are getting
-            # pushed to the url specified in the "SPACK_BUILDCACHE_COPY" environment
-            # variable.
-            src_url = remote_mirror_override or remote_mirror_url
-            dest_url = spack_buildcache_copy
-
-            stage_names.append("stage-copy-buildcache")
-            copy_job = {
-                "stage": "stage-copy-buildcache",
-                "tags": ["spack", "public", "medium", "aws", "x86_64"],
-                "image": "ghcr.io/spack/python-aws-bash:0.0.1",
-                "when": "on_success",
-                "interruptible": True,
-                "retry": service_job_retries,
-                "script": [
-                    ". ./share/spack/setup-env.sh",
-                    "spack --version",
-                    "aws s3 sync --exclude *index.json* --exclude *pgp* {0} {1}".format(
-                        src_url, dest_url
-                    ),
-                ],
-            }
-
-            output_object["copy-mirror"] = copy_job
-
         if rebuild_index_enabled:
             # Add a final job to regenerate the index
             stage_names.append("stage-rebuild-index")
@@ -1286,6 +1295,21 @@ def generate_gitlab_ci_yaml(
         if spack_stack_name:
             output_object["variables"]["SPACK_CI_STACK_NAME"] = spack_stack_name
 
+        if spack_buildcache_copy:
+            # Write out the file describing specs that should be copied
+            copy_specs_dir = os.path.join(pipeline_artifacts_dir, "specs_to_copy")
+
+            if not os.path.exists(copy_specs_dir):
+                os.makedirs(copy_specs_dir)
+
+            copy_specs_file = os.path.join(
+                copy_specs_dir,
+                "copy_{}_specs.json".format(spack_stack_name if spack_stack_name else "rebuilt"),
+            )
+
+            with open(copy_specs_file, "w") as fd:
+                fd.write(json.dumps(buildcache_copies))
+
         sorted_output = {}
         for output_key, output_value in sorted(output_object.items()):
             sorted_output[output_key] = output_value
diff --git a/lib/spack/spack/cmd/buildcache.py b/lib/spack/spack/cmd/buildcache.py
index 4c1e4d4837..6aaa5eb1c7 100644
--- a/lib/spack/spack/cmd/buildcache.py
+++ b/lib/spack/spack/cmd/buildcache.py
@@ -2,6 +2,8 @@
 # Spack Project Developers. See the top-level COPYRIGHT file for details.
 #
 # SPDX-License-Identifier: (Apache-2.0 OR MIT)
+import glob
+import json
 import os
 import shutil
 import sys
@@ -271,7 +273,12 @@ def setup_parser(subparser):
 
     # Sync buildcache entries from one mirror to another
     sync = subparsers.add_parser("sync", help=sync_fn.__doc__)
-    source = sync.add_mutually_exclusive_group(required=True)
+    sync.add_argument(
+        "--manifest-glob",
+        default=None,
+        help="A quoted glob pattern identifying copy manifest files",
+    )
+    source = sync.add_mutually_exclusive_group(required=False)
     source.add_argument(
         "--src-directory", metavar="DIRECTORY", type=str, help="Source mirror as a local file path"
     )
@@ -281,7 +288,7 @@ def setup_parser(subparser):
     source.add_argument(
         "--src-mirror-url", metavar="MIRROR_URL", type=str, help="URL of the source mirror"
     )
-    dest = sync.add_mutually_exclusive_group(required=True)
+    dest = sync.add_mutually_exclusive_group(required=False)
     dest.add_argument(
         "--dest-directory",
         metavar="DIRECTORY",
@@ -614,6 +621,31 @@ def copy_fn(args):
     shutil.copyfile(specfile_src_path_yaml, specfile_dest_path_yaml)
 
 
+def copy_buildcache_file(src_url, dest_url, local_path=None):
+    """Copy from source url to destination url"""
+    tmpdir = None
+
+    if not local_path:
+        tmpdir = tempfile.mkdtemp()
+        local_path = os.path.join(tmpdir, os.path.basename(src_url))
+
+    try:
+        temp_stage = Stage(src_url, path=os.path.dirname(local_path))
+        try:
+            temp_stage.create()
+            temp_stage.fetch()
+            web_util.push_to_url(local_path, dest_url, keep_original=True)
+        except web_util.FetchError as e:
+            # Expected, since we have to try all the possible extensions
+            tty.debug("no such file: {0}".format(src_url))
+            tty.debug(e)
+        finally:
+            temp_stage.destroy()
+    finally:
+        if tmpdir and os.path.exists(tmpdir):
+            shutil.rmtree(tmpdir)
+
+
 def sync_fn(args):
     """Syncs binaries (and associated metadata) from one mirror to another.
     Requires an active environment in order to know which specs to sync.
@@ -622,6 +654,10 @@ def sync_fn(args):
         src (str): Source mirror URL
         dest (str): Destination mirror URL
     """
+    if args.manifest_glob:
+        manifest_copy(glob.glob(args.manifest_glob))
+        return 0
+
     # Figure out the source mirror
     source_location = None
     if args.src_directory:
@@ -687,8 +723,9 @@ def sync_fn(args):
         buildcache_rel_paths.extend(
             [
                 os.path.join(build_cache_dir, bindist.tarball_path_name(s, ".spack")),
-                os.path.join(build_cache_dir, bindist.tarball_name(s, ".spec.yaml")),
+                os.path.join(build_cache_dir, bindist.tarball_name(s, ".spec.json.sig")),
                 os.path.join(build_cache_dir, bindist.tarball_name(s, ".spec.json")),
+                os.path.join(build_cache_dir, bindist.tarball_name(s, ".spec.yaml")),
             ]
         )
 
@@ -701,24 +738,31 @@ def sync_fn(args):
             dest_url = url_util.join(dest_mirror_url, rel_path)
 
             tty.debug("Copying {0} to {1} via {2}".format(src_url, dest_url, local_path))
-
-            stage = Stage(
-                src_url, name="temporary_file", path=os.path.dirname(local_path), keep=True
-            )
-
-            try:
-                stage.create()
-                stage.fetch()
-                web_util.push_to_url(local_path, dest_url, keep_original=True)
-            except web_util.FetchError as e:
-                tty.debug("spack buildcache unable to sync {0}".format(rel_path))
-                tty.debug(e)
-            finally:
-                stage.destroy()
+            copy_buildcache_file(src_url, dest_url, local_path=local_path)
     finally:
         shutil.rmtree(tmpdir)
 
 
+def manifest_copy(manifest_file_list):
+    """Read manifest files containing information about specific specs to copy
+    from source to destination, remove duplicates since any binary packge for
+    a given hash should be the same as any other, and copy all files specified
+    in the manifest files."""
+    deduped_manifest = {}
+
+    for manifest_path in manifest_file_list:
+        with open(manifest_path) as fd:
+            manifest = json.loads(fd.read())
+            for spec_hash, copy_list in manifest.items():
+                # Last duplicate hash wins
+                deduped_manifest[spec_hash] = copy_list
+
+    for spec_hash, copy_list in deduped_manifest.items():
+        for copy_file in copy_list:
+            tty.debug("copying {0} to {1}".format(copy_file["src"], copy_file["dest"]))
+            copy_buildcache_file(copy_file["src"], copy_file["dest"])
+
+
 def update_index(mirror_url, update_keys=False):
     mirror = spack.mirror.MirrorCollection().lookup(mirror_url)
     outdir = url_util.format(mirror.push_url)