UPnP handling by Transmission causes download stalls?

[kovalev]

There were reports in this forum that download in Transmission (version 1.76 and lower) suddenly gets stalled. I have occasionally experienced this issue too. After having run version 1.80b5, where dealing with trackers had been greatly improved (many thanks to the Transmission team!), I seems have found a coincidence.

My connection to the Internet is via a wireless router with UPnP enabled. Thus, I use an option "Use UPnP/NAT-PMP port forwarding from my router" from Network/Preferences which seems is working properly. Even though Transmission port test still reports my port as "closed" (may be due to standard settings of router's firewall), there is no problem with incoming connections, and popular web services like canyouseeme.org report Transmission port isn't blocked.

Sometimes all network transmissions get stalled for several seconds, or even network auto reconnection occurs. Even though not affecting other applications these events occasionally do leave download in Transmission stalled. It happens only to peers that had been connected (uploading to me) before disruption has occurred, and seems only when PNP_EVENT_AUTORENEWAL_FAILED with auto uuid re-assignment has followed:
...
2010-01-13 17:21:49: Universal Plug and Play: UPNP_EVENT_RECEIVED:
2010-01-13 17:21:49: SID: uuid:2c825b7a-1dd2-11b2-98a7-d4bd857046f0
2010-01-13 17:21:49: Key: 62
2010-01-13 17:21:49: Property list:
2010-01-13 17:21:49: s:PortMappingNumberOfEntries='2093128'
2010-01-13 17:21:52: Universal Plug and Play: UPNP_EVENT_RECEIVED:
2010-01-13 17:21:52: SID: uuid:2c825b7a-1dd2-11b2-98a7-d4bd857046f0
2010-01-13 17:21:52: Key: 63
2010-01-13 17:21:52: Property list:
2010-01-13 17:21:52: s:PortMappingNumberOfEntries='6647412'
2010-01-13 17:46:01: Universal Plug and Play: UPNP_EVENT_AUTORENEWAL_FAILED: Re-subscribed to EventURL 'http://192.168.1.254:80/upnp/event/igd/wanpppcInternet' with SID == 'uuid:8368bd6e-1dd2-11b2-98a7-d4bd857046f0'.
2010-01-13 17:46:01: Universal Plug and Play: Error: GetStateVariable("PortMappingNumberOfEntries"): in a call to UpnpGetServiceVarStatus: Error code :'404', Error description :'Invalid statevariable name'.
2010-01-13 17:46:02: Universal Plug and Play: UPNP_EVENT_RECEIVED:
2010-01-13 17:46:02: SID: uuid:8368bd6e-1dd2-11b2-98a7-d4bd857046f0
2010-01-13 17:46:02: Key: 0
2010-01-13 17:46:02: Property list:
2010-01-13 17:46:02: s:PossibleConnectionTypes='IP_Routed'
2010-01-13 17:46:02: s:ConnectionStatus='Unconfigured'
2010-01-13 17:46:02: s:ExternalIPAddress='***.***.***.***'
2010-01-13 17:46:02: s:PortMappingNumberOfEntries='6647412'
2010-01-13 17:46:19: Universal Plug and Play: UPNP_EVENT_RECEIVED:
2010-01-13 17:46:19: SID: uuid:8368bd6e-1dd2-11b2-98a7-d4bd857046f0
2010-01-13 17:46:19: Key: 1
2010-01-13 17:46:19: Property list:
2010-01-13 17:46:19: s:PortMappingNumberOfEntries='2093128'
2010-01-13 17:46:22: Universal Plug and Play: UPNP_EVENT_RECEIVED:
2010-01-13 17:46:22: SID: uuid:8368bd6e-1dd2-11b2-98a7-d4bd857046f0
2010-01-13 17:46:22: Key: 2
2010-01-13 17:46:22: Property list:
2010-01-13 17:46:22: s:PortMappingNumberOfEntries='6647412'
...
(Note: this is not a Transmission log, but download stall in Transmission always coincided with UPNP_EVENT_AUTORENEWAL_FAILED)

Neither a status change of affected peer nor disconnection occurs - a stalled peer is still present in Peers list with "D" status displayed, but its download rate drops to zero. The stall will last until next announcing to a tracker, which could take as long as an hour... However, downloading from a newly connected peer would proceed normally. Thus, downloading from old peers could be resumed too by bootstrapping from a saved list of peers or by requesting list of peers from a tracker manually or by pausing/resuming a stalled torrent (in 1.80b; 1.76 would require restarting the client). Interesting, an outgoing transfer seems is not affected: uploading to connected peers resumes immediately after disruption.

Thus, the question is: could such download stalls occur due to handling UPnP events by Transmission? Though the issue seems is not critical and maybe specific to only some hardware configuration, it could become pretty annoying when only a handful of seeders is available.

Now running Transmission 1.85b GTK r9924 on Ubuntu 9.10 (kernel 2.6.31.18 and Gnome 2.28.1).

[jabohn]

I'm getting the stalls too. Nothing will be downloading in transmission, then suddenly everything jumps and is downloading in huge amounts. Then at seemingly random intervals it stalls again. I am running the beta 4 for version 1.8 - just updated to the latest beta 5 now to see if there's any change.

[jabohn]

Thanks, I haven't noticed the stalls anymore since going to beta 5.

[kovalev]

It seems the issue of download stalls has been solved in Transmission releases 9959 and above. No more stalled peers after network disruption - download now goes on immediately as local network recovers. UPnP errors, when occurred, do not result in stalled peers too. What's more, the amount of bad downloaded chunks has been reduced significantly, and I couldn't find any error related to Transmission (directly or indirectly) in system logs. Thanks again to the Transmission team!

Now running Transmission 1.85b GTK r9963 on Ubuntu 9.10 (Linux kernel 2.6.31.18 and Gnome 2.28.1).