WIP: Move long-running (db) local torrents check to thread #7796
Conversation
qstokkink
requested review from
a team and
egbertbouman
and removed request for
a team
qstokkink
changed the title
qstokkink
changed the title
qstokkink
force-pushed
the
upd_chklocaltrts
branch
4 times, most recently
from
8212cdc to
8dffb8e
Compare
qstokkink
force-pushed
the
upd_chklocaltrts
branch
from
8dffb8e to
5d857e8
Compare
qstokkink
changed the title
| :type torrents: list[TorrentState] | ||
| """ | ||
| for t in torrents: | ||
| loop.call_soon_threadsafe(self.register_anonymous_task, f"Check health {t.infohash}", |
There was a problem hiding this comment.
I'm not sure if this is intentional or not, but I'm guessing check_torrent_health is still doing database stuff on the main thread (since that's where the only running eventloop is).
If it is intentional, you might as well keep the check functions on the main thread, and schedule the torrents_to_check/torrents_to_check_in_user_channel using run_threaded
There was a problem hiding this comment.
This seems like it would necessitate a major refactoring of the torrent checker to solve sanely then. Secondly, I think that this refactoring would overlap too much with #7328.
Taking that into account, I'll close this PR in favor of another (future) more thorough refactoring PR.
| """ | ||
| loop = loop or get_running_loop() | ||
| selected_torrents = self.torrents_to_check() |
There was a problem hiding this comment.
You may need to call db.disconnect after torrents_to_check/torrents_to_check_in_user_channel since they're not running on the main thread anymore.
qstokkink
changed the title
Fixes #7784 as discussed on the issue.
The
check_torrents_in_user_channelhas almost the exact same code ascheck_local_torrents. So, I also gave that the same treatment.Restarts on this PR:
PR / mac / build: build timed out.