diff --git a/scripts/find_missing_merge_requests.py b/scripts/find_missing_merge_requests.py
new file mode 100755
index 0000000000..09d3e9a581
--- /dev/null
+++ b/scripts/find_missing_merge_requests.py
@@ -0,0 +1,82 @@
+#!/usr/bin/env python3
+
+import click
+import multiprocessing
+import pathlib
+import requests
+import urllib.parse
+
+
+@click.command()
+@click.option('--token_path', type=str, default=pathlib.Path.home() / '.gitlab_token',
+              help='Path to text file with Gitlab token.')
+@click.option('--project_id', type=int, default=7107382,
+              help='Gitlab project id.')
+@click.option('--host', type=str, default='gitlab.com',
+              help='Gitlab host.')
+@click.option('--workers', type=int, default=10,
+              help='Number of parallel workers.')
+@click.option('--target_branch', type=str, default='master',
+              help='Merge request target branch.')
+@click.option('--begin_page', type=int, default=1,
+              help='Begin with given /merge_requests page.')
+@click.option('--end_page', type=int, default=4,
+              help='End before given /merge_requests page.')
+@click.option('--per_page', type=int, default=100,
+              help='Number of merge requests per page.')
+def main(token_path, project_id, host, workers, target_branch, begin_page, end_page, per_page):
+    token = read_token(token_path)
+    base_url = f'https://{host}/api/v4/projects/{project_id}/'
+    checked = 0
+    filtered = 0
+    missing = 0
+    for page in range(begin_page, end_page):
+        merge_requests = requests.get(
+            url=urllib.parse.urljoin(base_url, 'merge_requests'),
+            headers={'PRIVATE-TOKEN': token},
+            params=dict(state='merged', per_page=per_page, page=page),
+        ).json()
+        if not merge_requests:
+            break
+        checked += len(merge_requests)
+        merge_requests = [v for v in merge_requests if v['target_branch'] == target_branch]
+        if not merge_requests:
+            continue
+        filtered += len(merge_requests)
+        with multiprocessing.Pool(workers) as pool:
+            missing_merge_requests = pool.map(FilterMissingMergeRequest(token, base_url), merge_requests)
+            for mr in missing_merge_requests:
+                if mr is not None:
+                    missing += 1
+                    print(f"MR {mr['reference']} ({mr['id']}) is missing from branch {mr['target_branch']},"
+                          f" previously was merged as {mr['merge_commit_sha']}")
+    print(f'Checked {checked} MRs ({filtered} with {target_branch} target branch), {missing} are missing')
+
+
+class FilterMissingMergeRequest:
+    def __init__(self, token, base_url):
+        self.token = token
+        self.base_url = base_url
+
+    def __call__(self, merge_request):
+        commit_refs = requests.get(
+            url=urllib.parse.urljoin(self.base_url, f"repository/commits/{merge_request['merge_commit_sha']}/refs"),
+            headers={'PRIVATE-TOKEN': self.token},
+        ).json()
+        if 'message' in commit_refs and commit_refs['message'] == '404 Commit Not Found':
+            return merge_request
+        if not present_in_branch(commit_refs, branch=merge_request['target_branch']):
+            return merge_request
+
+
+def present_in_branch(commit_refs, branch):
+    return bool(next((v for v in commit_refs if v['type'] == 'branch' and v['name'] == branch), None))
+
+
+def read_token(path):
+    with open(path) as stream:
+        return stream.readline().strip()
+
+
+if __name__ == '__main__':
+    main()