Prevent project team from being truncated too early during project destruction
There are two issues with truncating the project team early: 1. `Projects::UnlinkForkService` may not close merge requests properly since permissions may be revoked early. 2. If an error is encountered during flushing of caches, then the user will lose all privileges, possibly causing an issue on deletion on retry.
This commit is contained in:
parent
12cd4c8360
commit
45f94ea78c
2 changed files with 20 additions and 2 deletions
|
@ -17,8 +17,6 @@ module Projects
|
|||
def execute
|
||||
return false unless can?(current_user, :remove_project, project)
|
||||
|
||||
project.team.truncate
|
||||
|
||||
repo_path = project.path_with_namespace
|
||||
wiki_path = repo_path + '.wiki'
|
||||
|
||||
|
@ -30,6 +28,7 @@ module Projects
|
|||
Projects::UnlinkForkService.new(project, current_user).execute
|
||||
|
||||
Project.transaction do
|
||||
project.team.truncate
|
||||
project.destroy!
|
||||
|
||||
unless remove_registry_tags
|
||||
|
|
|
@ -50,6 +50,25 @@ describe Projects::DestroyService, services: true do
|
|||
it { expect(Dir.exist?(remove_path)).to be_truthy }
|
||||
end
|
||||
|
||||
context 'when flushing caches fail' do
|
||||
before do
|
||||
new_user = create(:user)
|
||||
project.team.add_user(new_user, Gitlab::Access::DEVELOPER)
|
||||
allow_any_instance_of(Projects::DestroyService).to receive(:flush_caches).and_raise(Redis::CannotConnectError)
|
||||
end
|
||||
|
||||
it 'keeps project team intact upon an error' do
|
||||
Sidekiq::Testing.inline! do
|
||||
begin
|
||||
destroy_project(project, user, {})
|
||||
rescue Redis::CannotConnectError
|
||||
end
|
||||
end
|
||||
|
||||
expect(project.team.members.count).to eq 1
|
||||
end
|
||||
end
|
||||
|
||||
context 'with async_execute' do
|
||||
let(:async) { true }
|
||||
|
||||
|
|
Loading…
Reference in a new issue