Hi, When systemd tries to shutdown a service, after a certain timeout, it just kills all the processes belonging to this service. There is a configuration option that allows to change that timeout. However, it is available only for the new systemd formatI and not for the "legacy" (i.e. /etc/init.d/*) services.
The full story goes like this: transmission-daemon has an uncharacteristically long shutdown time. Every time the service goes down, systemd thinks that it is unresponsive and kills it. It messes up the metadata of the seeding (i.e. fully downloaded) torrents and when I bring the daemon back up, it thinks that it needs to re-download a torrent, even though the actual data is there. I have to manually select "Verify the local data" every time I start the daemon and it is incredibly annoying.
On Sun, 2011-06-05 at 09:52 +0300, Leonid Podolny wrote:
Hi, When systemd tries to shutdown a service, after a certain timeout, it just kills all the processes belonging to this service. There is a configuration option that allows to change that timeout. However, it is available only for the new systemd formatI and not for the "legacy" (i.e. /etc/init.d/*) services.
The full story goes like this: transmission-daemon has an uncharacteristically long shutdown time. Every time the service goes down, systemd thinks that it is unresponsive and kills it. It messes up the metadata of the seeding (i.e. fully downloaded) torrents and when I bring the daemon back up, it thinks that it needs to re-download a torrent, even though the actual data is there. I have to manually select "Verify the local data" every time I start the daemon and it is incredibly annoying.
I've had a somewhat similar experience with Qbittorrent, but it dates from F14 and so has nothing to do with systemd. The author tells me that the latest QBT fixes it (version 8, now in updates-testing). IMHO it's the responsibility of the app to store persistent data safely in anticipation of sudden exits or crashes, so I don't know that extending the timeout is really a solution.
poc
On Sun, 2011-06-05 at 09:52 +0300, Leonid Podolny wrote:
The full story goes like this: transmission-daemon has an uncharacteristically long shutdown time. Every time the service goes down, systemd thinks that it is unresponsive and kills it. It messes up the metadata of the seeding (i.e. fully downloaded) torrents and when I bring the daemon back up, it thinks that it needs to re-download a torrent, even though the actual data is there. I have to manually select "Verify the local data" every time I start the daemon and it is incredibly annoying.
Maybe you can add the verify step to the transmission-daemon init script. Or, better, you could create a wrapper script that starts the daemon and then does the verify.
I know, still shitty.
Also, this sucks because I just started using transmission-daemon instead of the GTK version in F12, and I've become rather fond of it! I was planning on setting it up on F15. Fail.
Regards,
Ranbir
-
Leonid Podolny -
Patrick O'Callaghan -
Ranbir