Test if remote repository exists before cloning

When a repository does not exist on a remote, Gitaly won't be able to
clone it. This is correct behaviour, but from the clients perspective a
change in behaviour.

This change implements the client side changes that allows Gitaly to
execute a `git ls-remote <remote-url> HEAD`. This way the client has no
need to shell out to Git.

In the situation where multiple Gitalies are available, one is chosen at
random.

This commit closes https://gitlab.com/gitlab-org/gitlab-ce/issues/43929,
while its also a part of https://gitlab.com/gitlab-org/gitaly/issues/1084

3 files changed, 21 insertions, 2 deletions

diff --git a/lib/gitlab/gitaly_client.rb b/lib/gitlab/gitaly_client.rb
index 8ca30ffc232..0abae70c443 100644
--- a/lib/gitlab/gitaly_client.rb
+++ b/lib/gitlab/gitaly_client.rb
@@ -83,6 +83,10 @@ module Gitlab
       end
     end
 
+    def self.random_storage
+      Gitlab.config.repositories.storages.keys.sample
+    end
+
     def self.address(storage)
       params = Gitlab.config.repositories.storages[storage]
       raise "storage not found: #{storage.inspect}" if params.nil?
diff --git a/lib/gitlab/gitaly_client/remote_service.rb b/lib/gitlab/gitaly_client/remote_service.rb
index 58c356edfd1..f2d699d9dfb 100644
--- a/lib/gitlab/gitaly_client/remote_service.rb
+++ b/lib/gitlab/gitaly_client/remote_service.rb
@@ -3,6 +3,17 @@ module Gitlab
     class RemoteService
       MAX_MSG_SIZE = 128.kilobytes.freeze
 
+      def self.exists?(remote_url)
+        request = Gitaly::FindRemoteRepositoryRequest.new(remote: remote_url)
+
+        response = GitalyClient.call(GitalyClient.random_storage,
+                                     :remote_service,
+                                     :find_remote_repository, request,
+                                     timeout: GitalyClient.medium_timeout)
+
+        response.exists
+      end
+
       def initialize(repository)
         @repository = repository
         @gitaly_repo = repository.gitaly_repository
diff --git a/lib/gitlab/github_import/importer/repository_importer.rb b/lib/gitlab/github_import/importer/repository_importer.rb
index ab0b751fe24..b1b283e98b5 100644
--- a/lib/gitlab/github_import/importer/repository_importer.rb
+++ b/lib/gitlab/github_import/importer/repository_importer.rb
@@ -16,7 +16,8 @@ module Gitlab
         # Returns true if we should import the wiki for the project.
         def import_wiki?
           client.repository(project.import_source)&.has_wiki &&
-            !project.wiki_repository_exists?
+            !project.wiki_repository_exists? &&
+            Gitlab::GitalyClient::RemoteService.exists?(wiki_url)
         end
 
         # Imports the repository data.
@@ -55,7 +56,6 @@ module Gitlab
 
         def import_wiki_repository
           wiki_path = "#{project.disk_path}.wiki"
-          wiki_url = project.import_url.sub(/\.git\z/, '.wiki.git')
           storage_path = project.repository_storage_path
 
           gitlab_shell.import_repository(storage_path, wiki_path, wiki_url)
@@ -70,6 +70,10 @@ module Gitlab
           end
         end
 
+        def wiki_url
+          project.import_url.sub(/\.git\z/, '.wiki.git')
+        end
+
         def update_clone_time
           project.update_column(:last_repository_updated_at, Time.zone.now)
         end