Transmission re-downloads stuff already completed?

[maki.ato]

This is a weird one.
Whenever I start transmission (either manually, or forced by a power out) transmission picks between 1 to 3 torrents that already completed and were seeding and downloads them from scratch. It shows their completing is at 0%. If I stop those downloads and verify them they go back to 100% and I can continue seeding as normal.
This started happening with v2.83. It also happens with 2.84.

HELP!

Technical details: I'm running transmission 2.84 on a D-Link DNS-323.

[maki.ato]

I guess my first post was "a bit" lacking in detail, so here goes: If it's unclear, I'm running Tranmsission Daemon on the DNS-323.
Uptime is 11 days (since the last power out, that's when I noticed this issue).

Transmission, its config file, log file, resume files, torrents, libs, compete/incomplete data all reside on the same partition on the same physical drive. The setup is simple, no RAID, JBOD or anything of the sort.
The drive is spun up, the partition is mounted.

I stopped tranmission, up-ed the message level to 3 (debug log level, previously it was 1) and restarted it. Sure enough, one complete torrent was at 0% and downloading. I immediately stopped it and verified it. Yep, 100% complete.
Here's all the log lines relevant to it:

Code: Select all

...
[2014-08-04 19:40:06.408 IDT] Couldn't save file "/var/lib/transmission/resume/Some Torrent.55023b328987ef73.resume": No such file or directory (variant.c:1219)
[2014-08-04 19:40:07.410 IDT] Some Torrent Unable to save resume file: No such file or directory (torrent.c:536)
[2014-08-04 19:40:07.411 IDT] Couldn't save file "/var/lib/transmission/resume/Some Torrent.55023b328987ef73.resume": No such file or directory (variant.c:1219)
[2014-08-04 19:40:07.412 IDT] Some Torrent Unable to save resume file: No such file or directory (torrent.c:536)
...
[2014-08-05 11:46:13.916 IDT] Couldn't read "/var/lib/transmission/resume/Some Torrent.55023b328987ef73.resume": Input/output error (utils.c:233)
[2014-08-05 11:46:13.917 IDT] Some Torrent Couldn't read "/var/lib/transmission/resume/Some Torrent.55023b328987ef73.resume" (resume.c:715)
...
[2014-08-05 11:46:33.420 IDT] Some Torrent Announcing to tracker (announcer.c:1552)
...
[2014-08-05 11:46:33.442 IDT] Some Torrent Got 9 peers from tracker (torrent.c:564)
...
[2014-08-05 11:46:42.458 IDT] Some Torrent Scrape successful. Rescraping in 1800 seconds. (announcer.c:1372)
...
[2014-08-05 11:46:59.443 IDT] Some Torrent [LAZY] checking just-completed piece 822 (torrent.c:3250)
[2014-08-05 11:47:01.539 IDT] Some Torrent Pausing (torrent.c:1857)
[2014-08-05 11:47:01.540 IDT] Saved "/var/lib/transmission/resume/Some Torrent.55023b328987ef73.resume" (variant.c:1214)
...
[2014-08-05 11:47:01.541 IDT] Some Torrent Announcing to tracker (announcer.c:1552)
...

[maki.ato]

More details:
It's probably not a HDD failure. I ran a SMART quick scan on the drive and all's well. "Completed without error".
In the previous log it seems that the issue has something to do with saving the resume file. That part of the log is from when the message level was set to 1. Now that it's 3, it's time to stop/start again and see what happens.
After stopping, I don't see in the log any new resume file saving, and when I start transmission daemon again all's good.

I do, however, see a lot of messages (removing, pausing). Maybe that higher log level is changing the timing on the shutdown process and prevents this issue from happening?
Maybe I should've added that I'm seeding ~850 torrents.

I'm trying again to stop/start and all's good. Previously, on every stop/start I'd get the issue. I'm starting to think that this debug message level is interfering in some way.

[maki.ato]

Take 3:
I stop transmission daemon (it was running with message level 3). When I check the log there's no new resume file saving errors. I change the message level back to 1 and start it.
All is good, no re-downloading already completed torrents.

I'm wait for a bit to let everything settle down before shutting transmission daemon down again.
Stop/start - All's still good.

[maki.ato]

That might explain the first time it happened after that power out. It doesn't explain the 3-4 times it happened afterwards when I stopped/started the daemon or rebooted the NAS (rebooting unmounts partitions cleanly). The partition, btw, is ext3 and it's journaled.

It can't be just that blackout. Transmission daemon had no problem with that "Some Torrent" resume file (please see that log excerpt I posted before) when it started/stopped several times since.

[maki.ato]

rebooted the NAS (rebooting unmounts partitions cleanly)

You did quit Transmission first?

Yes. And as I wrote before, even if I stop/start the daemon this happens.

[cfpp2p]

It doesn't sound good "power out". Try running a file system check, and with the power out problems this means all of your drives perhaps.

http://linux.about.com/od/commands/l/blcmdl8_fsck_.htm

http://www.linuxquestions.org/questions ... ost2832256

Systems aren't very good with power out situations. I would suggest an uninterrupted power supply.

[maki.ato]

I know about fsck, thanks. I have plenty of experience with Linux.
As I wrote before, and I hope if you'd read through it you'll agree, this power out does not explain why this issue repeated itself multiple times, both after clean system reboots and Transmission Daemon stop/start.

[maki.ato]

Now my nightly rsync snapshot job reports I/O errors with some of those resume files.
Looks like you were right all along. It's HDD related.

Thanks for all your help.