Pros:
- Using native protocol over executing binaries is always better
- `clickhouse-client` in debug build takes almost a second to execute simple `SELECT 1`
and `clickhouse-test` requires ~5 queries at start (determine some
flags, zk, alive, create database)
Notes:
- `FORMAT Vertical` had been replaced with printing of `pandas.DataFrame`
And after this patch tiny tests work with the speed of the test, and
does not requires +-5 seconds of bootstrapping.
Right now it is possible to get the following error:
Having 20 errors! 0 tests passed. 0 tests skipped. 57.37 s elapsed (MainProcess).
Won't run stateful tests because test data wasn't loaded.
Traceback (most recent call last):
File "/usr/lib/python3.9/multiprocessing/managers.py", line 802, in _callmethod
conn = self._tls.connection
AttributeError: 'ForkAwareLocal' object has no attribute 'connection'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/ch/clickhouse/.cmake/../tests/clickhouse-test", line 1462, in <module>
main(args)
File "/src/ch/clickhouse/.cmake/../tests/clickhouse-test", line 1261, in main
if len(restarted_tests) > 0:
File "<string>", line 2, in __len__
File "/usr/lib/python3.9/multiprocessing/managers.py", line 806, in _callmethod
self._connect()
File "/usr/lib/python3.9/multiprocessing/managers.py", line 793, in _connect
conn = self._Client(self._token.address, authkey=self._authkey)
File "/usr/lib/python3.9/multiprocessing/connection.py", line 507, in Client
c = SocketClient(address)
File "/usr/lib/python3.9/multiprocessing/connection.py", line 635, in SocketClient
s.connect(address)
ConnectionRefusedError: [Errno 111] Connection refused
The reason behind this is that manager's thread got terminated:
ipdb> p restarted_tests._manager._process
<ForkProcess name='SyncManager-1' pid=25125 parent=24939 stopped exitcode=-SIGTERM>
Refs: #29259 (cc: @vdimir)
Follow-up for: #29197 (cc: @tavplubix)
Variables aren't shared when using multiprocessing, use shared memory
instead
https://docs.python.org/3/library/multiprocessing.html#shared-ctypes-objects.
There appears to be a deadlock when multiple threads try to send
sigterm signal at the same time. Avoid it by making sure sigterm is sent
only once for the process group.
