1 |
Tom <uebershark@××××××××××.com> posted |
2 |
20090415143440.6553d0dd@ViciousVincent, excerpted below, on Wed, 15 Apr |
3 |
2009 14:34:40 +0200: |
4 |
|
5 |
> Hi, I've been using transmission as my main bittorrent client for ages |
6 |
> now. |
7 |
> |
8 |
> For some odd reason, it randomly just pauses all downloads. It normally |
9 |
> would do this, when running out of disk space, which is not the case |
10 |
> here.. |
11 |
> |
12 |
> Any ideas? |
13 |
|
14 |
This one's kind of random, but who knows? |
15 |
|
16 |
This is based on LWN's "Notes from the LSF storage track" article as |
17 |
found at the link below. More specifically, it's based on the first |
18 |
paragraph under ATA issues, about the problem with zero-length barriers. |
19 |
Apparently this causes some drives to power off briefly, possibly due to |
20 |
power-saving then powering up for the flush, drawing more power than |
21 |
expected. |
22 |
|
23 |
http://lwn.net/Articles/328347/ |
24 |
|
25 |
It's causing problems with occasional lost data due to losing the cache |
26 |
in the power blink. Naturally, this confuses the drivers and possibly |
27 |
the apps using the drive as well, particularly if an error is propagated |
28 |
up the driver stack to them during the momentary outage. |
29 |
|
30 |
I wouldn't consider this very likely except that I just read it |
31 |
yesterday, so it was top of my mind when I read your problem. It's |
32 |
possible, particularly if the problem is relatively recent, with newer |
33 |
kernels. I'm not sure when those barriers were enabled and there's not a |
34 |
lot more info in the article than I posted, but the article does imply |
35 |
it's a fairly recently encountered problem. |
36 |
|
37 |
More generally, it could be any similar issue that causes a temporary |
38 |
error. There's a few temporary errors that could, if taken wrong by some |
39 |
layer of the stack (not necessarily transmission itself), cause the app |
40 |
to pause, thinking the storage is full, has been removed, or some other |
41 |
such thing. |
42 |
|
43 |
BTW, I don't use transmission (I don't do much torrenting and I'm |
44 |
normally a KDE guy so what I've done has been with ktorrent), but if you |
45 |
take a look at my headers, you'll see I'm using pan. Pan's lead dev is |
46 |
Charles Kerr, tho he hasn't done a lot with it for nearing two years |
47 |
now. However, during at least part of that time I've heard he's been |
48 |
working on transmission. =:^) It had apparently been abandoned or nearly |
49 |
so for some time, until Charles came along to work on it (that's only |
50 |
what I read, I don't know, no offense to anyone intended if it's a group |
51 |
effort). Meanwhile, I've been quite active for years as a regular on the |
52 |
pan mailing lists (newsgroups as served thru gmane, just like this list/ |
53 |
group, where I read and reply with pan) and believe I may now be the |
54 |
senior regular there. Thus, even tho I don't use transmission, I feel |
55 |
like I have a bit of a connection with anyone that does, since the |
56 |
project I've been a long time list regular for has as the primary dev one |
57 |
of the guys who has put a lot of work into transmission now as well. =:^) |
58 |
|
59 |
So anyway, see if it has a mailing list. If it does and it's as active |
60 |
as the pan user list, you'll likely find help. |
61 |
|
62 |
Meanwhile, one other possibility. If you run GNOME, see if you have the |
63 |
Assistive Technologies daemon (should be a process named at-spi-registryd) |
64 |
running. With pan it has been a problem, most often seen on Ubuntu since |
65 |
they apparently have it running by default now, but it could be a problem |
66 |
on other distributions if it's running, too. Apparently, it polls the |
67 |
input and other devices very frequently, to the point it slows down a |
68 |
bunch of other stuff including pan downloading. We tried and tried to |
69 |
figure that one out, finding out it was Ubuntu only (for the most part), |
70 |
then Ubuntu, only when running GNOME, which was strange as pan is a gtk/ |
71 |
gnome family app, then we were stuck for awhile and could only tell |
72 |
people to try KDE or XFCE when running pan. But someone figured it out |
73 |
recently, apparently as he'd had problems with it sucking performance out |
74 |
of other things as well. |
75 |
|
76 |
So anyway, if you have an at-spi-registryd process running, consider |
77 |
killing it unless you really need it due to being "alternately enabled". |
78 |
See if that helps. On Ubuntu-8.10/Gnome at least, it's under System > |
79 |
Preferences > Assistive Technologies. The instructions as posted on the |
80 |
pan list/group say uncheck "Enable Assistive Technologies". The poster |
81 |
said he rebooted soon after and the problem was gone, but he didn't know |
82 |
if a reboot was required. |
83 |
|
84 |
-- |
85 |
Duncan - List replies preferred. No HTML msgs. |
86 |
"Every nonfree program has a lord, a master -- |
87 |
and if you use the program, he is your master." Richard Stallman |