Downloads slow, stop, get stuck on "starting up"

[Jay S]

Binary package hint: miro

Encountered this problem in both Miro 3.0.3 and 3.5, under Ubuntu Maverick.

Under Lucid, everything was working fine. Now, downloads begin, pick up speed, then start to slow until they're at a couple bytes per second, before finally stalling at "starting up" and no further progress is made.

I have adequate disk space; my firewall is turned off; ports are forwarded; and the same torrent works in Transmission.

I uninstalled Miro, deleted my ~/.miro directory, reinstalled, and encountered the same problem.

[David Jensen (david-lgj)]

I have exactly the same problem with Miro 3.0.3.

It worked just fine until I upgraded to Ubuntu Maverick a couple of days ago.
The only thing I've found that looks a bit suspicious in Miros output is this:

2010-11-01 13:40:14,075 INFO Movie data process hung, killing it
2010-11-01 13:40:15,085 INFO Movie data process killed

[Jay S (topdownjimmy)]

Also of note:

libtorrent: 0.15.4.0
pycurl: libcurl/7.21.0 GnuTLS/2.8.6 zlib/1.2.3.4 libidn/1.18

[Jay S (topdownjimmy)]

I believe this same problem is being described at Miro's Bugzilla:

http://bugzilla.pculture.org/show_bug.cgi?id=12381

[Bryce Harrington (bryce)]

I seem to be able to reproduce this. I started a download set to a limit of 4 KB/s max. Initially it indicated downloading at a rate of about 3 kb/s, then after a minute or so this dropped to around 2 kb/s and now after about 20 min it's down to 1 kb/s. I haven't seen it get "stuck" though, but the rate is still decreasing little by little so I'll believe it might get to that point.

This was while downloading the "Pioneer One From VODO Promoted Works", which was the first item listed in Miro after a fresh installation.

It seems to me if I could reproduce it this easily it may be a widespread issue, and since downloading things is a pretty fundamental functionality of Miro I'm setting this to High importance.

[janet (jed-pculture)]

Miro bug: http://bugzilla.pculture.org/show_bug.cgi?id=12381 - was marked
as a duplicate of Miro bug
http://bugzilla.pculture.org/show_bug.cgi?id=13549 -

it appears to be an issue specific to libtorrent 0.15.x.

Marking bug http://bugzilla.pculture.org/show_bug.cgi?id=13549 a p1 for miro
4.0.

On Wed, Dec 15, 2010 at 8:10 PM, Bryce Harrington <<email address hidden>
> wrote:

> ** Also affects: democracy via
> http://bugzilla.pculture.org/show_bug.cgi?id=12381
> Importance: Unknown
> Status: Unknown
>
> --
> You received this bug notification because you are subscribed to Miro.
> https://bugs.launchpad.net/bugs/668615
>
> Title:
> Downloads slow, stop, get stuck on "starting up"
>

[Jay S (topdownjimmy)]

This appears to have been a problem with libtorrent-rasterbar 0.15.4
https://launchpad.net/ubuntu/+source/libtorrent-rasterbar

It was fixed in libtorrent-rasterbar 0.15.5, as reported in this bug on Miro's Bugzilla:
http://bugzilla.pculture.org/show_bug.cgi?id=13549

You can find the new version of the libtorrent-rasterbar package in the qBittorrent PPA:
https://launchpad.net/~hydr0g3n/+archive/ppa

Add the PPA to your software sources, update the package, and the problem should be solved. Finally! :)

[Martin Pitt (pitti)]

Moving to correct source package then, and unassigning from desktop team as this is an universe package. I subscribe sponsors instead, to apply the upstream fix to the package (or update to a newer version if it's already released).

[Sebastien Bacher (seb128)]

comment #15 states the issue is fixed in 0.15.5 which is the version in natty

[Michael Vogt (mvo)]

I just looked into this from the perspective of sposnoring a SRU for maverick. I am not able to see a isolated patch for the problem. I unsubscribe ubuntu-sponsors for now. If the fix from the Changes (below) can be isolated, please resubscribe ubuntu-sponsors.

+0.15.5 release
+
+ * support DHT extension to report external IPs
+ * fixed rare crash in http_connection's error handling
+ * avoid connecting to peers listening on ports < 1024
+ * optimized piece picking to not cause busy loops in some end-game modes
+ * fixed python bindings for tcp::endpoint
+ * fixed edge case of pad file support
+ * limit number of torrents tracked by DHT
+ * fixed bug when allow_multiple_connections_per_ip was enabled
+ * potential WOW64 fix for unbuffered I/O (windows)
+ * expose set_alert_queue_size_limit to python binding
+ * support dht nodes in magnet links
+ * support 100 Continue HTTP responses
+ * changed default choker behavior to use 8 unchoke slots (instead of bei
ng rate based)
+ * fixed error reporting issue in disk I/O thread
+ * fixed file allocation issues on linux
+ * fixed filename encoding and decoding issue on platforms using iconv
+ * reports redundant downloads to tracker, fixed downloaded calculation t
o
+ be more stable when not including redundant. Improved redundant data a
ccounting
+ to be more accurate
+ * fixed bugs in http seed connection and added unit test for it
+ * fixed error reporting when fallocate fails
+ * deprecate support for separate proxies for separate kinds of connectio
ns

[JC Hulce (soaringsky)]

This bug affects Ubuntu 10.10, Maverick Meerkat. Maverick has reached end-of-life and is no longer supported, so I am closing the bugtask for Maverick. Please upgrade to a newer version of Ubuntu.
More information here: https://lists.ubuntu.com/archives/ubuntu-announce/2012-April/000158.html