Commit 4b34720c authored by Bob Van Landuyt's avatar Bob Van Landuyt

Use ruby's `File.stat` to check storage availability

parent 3a2aaed8
......@@ -11,7 +11,7 @@ module Gitlab
end
def timeout_check(path, timeout_seconds)
filesystem_check_pid = check_filesystem_in_fork(path)
filesystem_check_pid = check_filesystem_in_process(path)
deadline = timeout_seconds.seconds.from_now.utc
wait_time = 0.01
......@@ -31,27 +31,23 @@ module Gitlab
status
end
# This call forks out into a process, that process will then be replaced
# With an `exec` call, since we fork out into a shell, we can create a
# child process without needing an ActiveRecord-connection.
# This will spawn a new 2 processes to do the check:
# The outer child (waiter) will spawn another child process (stater).
#
# Inside the shell, we use `& wait` to fork another child. We do this
# to prevent leaving a zombie process when the parent gets killed by the
# timeout.
#
# https://stackoverflow.com/questions/27892975/what-causes-activerecord-breaking-postgres-connection-after-forking
# https://stackoverflow.com/questions/22012943/activerecordstatementinvalid-runtimeerror-the-connection-cannot-be-reused-in
def check_filesystem_in_fork(path)
fork do
STDOUT.reopen('/dev/null')
STDERR.reopen('/dev/null')
exec("(#{test_script(path)}) & wait %1")
end
end
def test_script(path)
"testpath=$(realpath #{Shellwords.escape(path)}) && stat \"$testpath\""
# The stater is the process is performing the actual filesystem check
# the check might hang if the filesystem is acting up.
# In this case we will send a `KILL` to the waiter, which will still
# be responsive while the stater is hanging.
def check_filesystem_in_process(path)
spawn('ruby', '-e', ruby_check, path, [:out, :err] => '/dev/null')
end
def ruby_check
<<~RUBY_FILESYSTEM_CHECK
inner_pid = fork { File.stat(ARGV.first) }
Process.waitpid(inner_pid)
exit $?.exitstatus
RUBY_FILESYSTEM_CHECK
end
end
end
......
......@@ -17,11 +17,20 @@ describe Gitlab::Git::Storage::ForkedStorageCheck, skip_database_cleaner: true d
end
it 'returns false when the check takes to long' do
allow(described_class).to receive(:check_filesystem_in_fork) do
fork { sleep 10 }
# We're forking a process here that takes too long
# It will be killed it's parent process will be killed by it's parent
# and waited for inside `Gitlab::Git::Storage::ForkedStorageCheck.timeout_check`
allow(described_class).to receive(:check_filesystem_in_process) do
Process.spawn("sleep 10")
end
result = true
expect(described_class.storage_available?(existing_path, 0.5)).to be_falsey
runtime = Benchmark.realtime do
result = described_class.storage_available?(existing_path, 0.5)
end
expect(result).to be_falsey
expect(runtime).to be < 1.0
end
describe 'when using paths with spaces' do
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment