Tumbleweed from you guys lol ....... Testing Tixati again today for this issue. Thought you might be interested in the results...
This problem seems to be getting
WORSE and WORSE with Transmission peers.
This is getting seriously annoying.
I have no idea why other people are not seeing this? This cannot just be "my problem"
I am encountering this issue with about a 15% reproduction rate.
After spinning up some virtual machines today, I've installed various OS's, various Torrent client programs and I've tested dumping a list of hash-links for hours.
The hash-links chosen are random, lesser seeded, but they ARE definitely seeded.
Conclusion: This is NOT a Tixati bug. This is entirely a software bug in Transmission side. A really horrendous one as well.
Indeed this SOFTWARE BUG is genuine and reproducible in a Data Center between VM's Public IP's on quite different subnets.
EXAMPLE seen in Tixati
Test Conditions:
01) Tixati (and other clients each separately spun up differently) each only have knowledge of the hash-link/magnet link (no files, no .torrent)
02) Availability online is 1 Transmission full SEED
03) Transmission SEED has 100% file availability (some public, 1x I am hosting for testing)
04) Transmission v2, v3, v4 even latest v4 release affected
05) Problem is ONLY visible when Transmission tries to "serve" its Metadata to any peer with zero Metadata.
06) If you give any third party peer the Metadata (.torrent) it will download the file from Transmission absolutely perfectly. So the machine is not a spoof or broken. Downloaded Files confirmed genuine.
07) Rebooting Transmission or the Operating System DOES NOT fix the problem with Transmission when self-hosted with bug condition.
08) Changing Transmission ports/settings DOES NOT fix the problem with it refusing to serve the Metadata out to Tixati etc
09) hash-links / Metadata affected are random and I do not know why
10) Metadata is broken into "pieces"... Let's say 5 pieces (index=0 to 4?). All Metadata pieces download APART FROM 1 PIECE, so only 80% of Metadata is obtained. Not enough to start download. (Client aborts)
FORCE FIX 1
11) I can force fix the issue giving ANOTHER brand of client the Metadata (in form of .torrent file)
a) Then (qbit/utor) downloads the files perfectly OK from Transmission.
b) qbit/utor will serve Tixati the Metadata - that Transmission refuses to
c) Tixati will download the Metadata from qbit/utor and the files from Transmission as perfectly normal now.
Essentially this bug condition is only prevalent when Transmission is the only SEED and nobody can get started.
TIXATI LOG:
Connecting to Transmission with METADATA BUG CONDITION = true
[12:01:07] log detail level set to 4
[12:01:26] starting
[12:01:26] initiating connection
[12:01:26] securing connection
[12:01:27] sent handshake
[12:01:27] logged in
[12:01:27] sent bitfield
[12:01:28] remote is partial seed <---------- Transmission peer (mentioned here) is actually 100% complete.
[12:01:28] received PEX message
[12:01:28] requested metadata piece 2
[12:01:30] remote rejected request for metadata at index 2
[12:01:35] remote unchoked local
[12:01:51] requested metadata piece 2
[12:01:55] remote rejected request for metadata at index 2
[12:02:16] requested metadata piece 2
[12:02:19] remote rejected request for metadata at index 2
[12:02:40] requested metadata piece 2
[12:02:42] remote rejected request for metadata at index 2
[12:03:03] requested metadata piece 2
[12:03:06] remote rejected request for metadata at index 2
[12:03:27] requested metadata piece 2
[12:03:29] remote rejected request for metadata at index 2
Above shows Transmission REFUSING to serve one tiny piece of the Metadata. This is a really horrendous bug and I'm seeing it in ALL versions of Transmission including latest build.
Thought you might be interested that this issue exists. Anyone who experiences "Waiting for Metadata..." from a Transmission Seed - please add to the thread...