I have a strange performance problem that I can't make heads and tails of. Maybe anybody have any ideas,?
What I see is that shortly after I start Transmission I get decent DL speed and UL speeds, approx 1Mb+ cs 80-90 kb/sec. After a few hours or after the first files have completed to DL the DL speed drops to 20-50 kb/s and the UL speed drops to 2-6 kbits/s. Stopping Transmission and restarting it usually but not always brings the speed up again.
Som facts: - I've got a NV+ with 1Gb RAM, RAIDiator 4.1.8 and Transmission 2.42. I've got plenty of space available on the NAS, even on the hdc1. - My Internet connection is a 24/1 Mbit xDSL line and I've checked so I know I get at least 13/1 which would allow me to DL with roughly 1.3-1.4 Mb/sec but atleast 1Mb/sec. - I do not have any speedlimitations set - The NAS is connected to a WNDR3800 thru a 1Gbit/s Netgear switch so I have plenty of speed avaialble there - I've even done tests with only the NAS and my laptop connected (ie. disconnected the AppleTV 2:nd Gen, the PS3, the AirportExpress, my Server, our TV, the HD-box, my wifes laptop and my daughters laptop). - I've gone thru the motions with changing the settings.json parameters for cache, peers, not using the default BT ports, etc. so I'm quite sure this is not the problem, although if someone can share their settings I'm interested. - I have tried different versions of Transmission but either they have the same problem or they have other worse issues, like crashing. - I've tried the built in BT client which provides much better speed but much less functionality so I'm pretty sure it's not my ISP or the NAS that is the limitation and I'd rather run Transmission if I can since it is more feature rich.
If your router does not forward the right port to your NAS, you can not get new incoming connections. So, for as long as you are asking the tracker for new peers, you can download fast and you will upload to the peers that you are downloading from. However, incoming connections get blocked, so others will not connect to you. Eventually, this reduces the amount of peers you are connected to and with that, your download and upload speed.
You can check if you are forwarding the right port by going to the webinterface of Transmission and look at the Network-tab of the preference window. It should state if the port you've set is open or not. If your router supports upnp (and has it turned on), you can enable the option "use port forwarding from my router" to automatically forward the right port at startup.
That does seem like the problem but .. - I've got uPNP enabled in the WNDR3800 - I have the setting "use port forwarding ..." set for Transmission - I'm randomizing the port but I've also tried with different fixed ports - I've got the same ports specified in settings.json (60000-60999) and opened in the router - Fixed or Random, all these ports are all sent to the NAS and both for UTP and TCP - In the Router logs I can see that that the Portforwarding is done so I'm pretty sure this works - On the NAS with Netstat -an | grep -i [portnumber] I can see that there are a lot of connections on the current portnumber
Regardless of this the WebGUI says "Status: Unknown" for the portforwarding and the speeds are as previously described.
Feels like I'm either missing something very simple but I can't see it or this is a very odd and very specific problem that "only" I have.
So, any other ideas? I'm open to any help and any suggestions.
if the menu of the WNDR3800 is something like the one on my WNR2200, then there is a page with the current upnp ports. Can you check if the port assigned by Transmission is set to the right IP-address. It might happen that it registers the port-forwarding to the wrong IP, for instance that of the ReadyNAS Remote-IP (if you have that running).
This should not be a problem if you turn upnp off in Transmission and manually set the port in Transmission to a fixed one and forward that port to your NAS's IP.
What strikes me as odd is that the Router seems to be simply creating a new port for every time I start Transmission and leaving the old port opened. At least I assume this is what I see since I can see the "old" static default port 51413 which I haven't used for a few weeks now. Is this the way it should look like? Or isn't it not one line per translation and not per connection, or? What I mean is if I look at the NAS with "netstat -an | grep :60*" I see hundreds of connections. Or how does this really work?
But I can also with " netstat -an | grep x.x | grep EST | grep -v .x.y| grep -v :60221" see a lot of other ports that are in use by the NAS. Some I can understand why they are there (SSH, CIFS/SMB, DLNA, etc ... ) but some I don't know what they are there for. Maybe they are used by Transmission, or?
tcp 0 0 192.168.x.x:1535 82.238.118.184:51413 ESTABLISHED tcp 0 0 192.168.x.x:1957 175.136.7.46:42176 ESTABLISHED tcp 0 0 192.168.x.x:4248 94.75.78.110:51413 ESTABLISHED tcp 1439 0 192.168.x.x:4933 68.47.253.116:18480 ESTABLISHED tcp 0 0 192.168.x.x:2845 109.203.53.65:21766 ESTABLISHED tcp 0 0 192.168.x.x:4395 93.82.143.2:40041 ESTABLISHED tcp 15653 0 192.168.x.x:1721 90.224.93.54:44801 ESTABLISHED tcp 0 0 192.168.x.x:1082 199.59.241.239:80 ESTABLISHED tcp 0 0 192.168.x.x:1267 70.232.56.185:55473 ESTABLISHED tcp 0 0 192.168.x.x:3821 68.63.135.129:14607 ESTABLISHED tcp 5 0 192.168.x.x:2131 24.205.85.4:56698 ESTABLISHED tcp 0 0 192.168.x.x:1186 86.16.189.181:43309 ESTABLISHED tcp 1024 0 192.168.x.x:2312 85.247.197.146:32422 ESTABLISHED tcp 0 0 192.168.x.x:1429 124.178.34.125:47199 ESTABLISHED tcp 0 0 192.168.x.x:1132 220.4.91.95:21871 ESTABLISHED tcp 14983 0 192.168.x.x:2410 109.108.198.125:56720 ESTABLISHED tcp 16060 0 192.168.x.x:4599 145.97.217.230:58967 ESTABLISHED tcp 0 13880 192.168.x.x:2331 109.58.20.31:54555 ESTABLISHED tcp 6000 0 192.168.x.x:2371 85.64.11.98:51414 ESTABLISHED tcp 13032 0 192.168.x.x:2343 93.152.143.137:6954 ESTABLISHED tcp 0 0 192.168.x.x:2107 78.101.10.247:62205 ESTABLISHED tcp 0 0 192.168.x.x:3427 79.103.253.120:50139 ESTABLISHED tcp 0 0 192.168.x.x:4277 85.74.42.181:16042 ESTABLISHED tcp 32794 0 192.168.x.x:1357 176.31.108.31:23880 ESTABLISHED tcp 0 0 192.168.x.x:3309 71.237.114.248:36654 ESTABLISHED tcp 0 344 192.168.x.x:4086 95.211.88.54:80 ESTABLISHED tcp 6157 0 192.168.x.x:1440 188.165.198.56:16301 ESTABLISHED tcp 0 0 192.168.x.x:4097 85.247.219.171:51413 ESTABLISHED tcp 0 0 192.168.x.x:3802 66.8.164.137:3579 ESTABLISHED tcp 0 0 192.168.x.x:2998 129.241.124.236:28476 ESTABLISHED tcp 229558 0 192.168.x.x:2948 219.222.21.212:11916 ESTABLISHED tcp 0 0 192.168.x.x:2692 88.235.103.134:11114 ESTABLISHED tcp 0 0 192.168.x.x:2324 113.142.11.113:2262 ESTABLISHED tcp 0 0 192.168.x.x:3023 80.108.91.227:39590 ESTABLISHED tcp 0 0 192.168.x.x:1648 94.8.10.220:11418 ESTABLISHED tcp 0 0 192.168.x.x:3219 178.76.158.61:39873 ESTABLISHED tcp 0 0 192.168.x.x:4570 2.90.197.96:48090 ESTABLISHED tcp 186 0 192.168.x.x:1306 68.81.165.49:33341 ESTABLISHED tcp 0 14520 192.168.x.x:4533 114.91.147.194:55390 ESTABLISHED tcp 0 0 192.168.x.x:3890 64.180.149.144:53649 ESTABLISHED tcp 0 0 192.168.x.x:2459 115.64.89.225:23762 ESTABLISHED tcp 51 0 192.168.x.x:4070 142.177.75.192:50320 ESTABLISHED tcp 0 0 192.168.x.x:4973 80.212.153.56:13862 ESTABLISHED
Another thing that might be worth mentioning is what I see when I look at the connections using "netstat -an | [a lot of greps to includes and excludes to filter what I want to see] | wc -l". What I exclude is basically all the local connections and to the location where I am now and what I include is basically everything. So what I then end up with is: TOTAL: 398 ESTABLISHED: 82 SYN_RECV: 89 CLOSE_WAIT: 133 LAST_ACK: 75
If I do the same but add to the filter at the end to filter on Transmission port as well I see for the Transmission port: TOTAL : 270 ESTABLISHED: 19 SYN_RECV: 86 CLOSE_WAIT: 98 LAST_ACK: 64
The time between these values are the time it takes to execute the script that I built for this, so it is more or less at the same seconds that all these values are for.
I think the connections to port 60221 (2nd code block) are the peers connected to you and the other ones (3rd code block) are either to other services (CIFS,HTTP,etc) or are the connections you made to other people to download from.
Your router log shows that it is forwarding the right port to the right IP, so that shouldn't be a problem. I've got no idea why it doesn't delete the unused upnp ports. It also shows that you're connected to plenty peers, so that shouldn't be a problem either, which means I'm kinda out of options...
So, for as long as you are asking the tracker for new peers, you can download fast and you will upload to the peers that you are downloading from. However, incoming connections get blocked, so others will not connect to you. Eventually, this reduces the amount of peers you are connected to and with that, your download and upload speed.
The more I think about the more I think this is the problem.
I wish I could get a better logging from Transmission so that I can understand what it is doing and more importantly, what it is not doing. I've search but I can't seem to find a way to turn on better logging. Do you know of any way to do it?
I'm assuming Transmission is keeping some kind of log somewhere, but my knowledge (or lack thereof) of Linux tells me not to enable SSH so I can't go look for it.
What you can try is disabling upnp on your router and enabling a DMZ for your NAS. That way all ports are forwarded to your NAS. If that would give you better downloading, then there is most likely some port-forwarding problem (or another problem in your router).
I think Transmission is supposed to save everything in /var/log/daemon.log but I'm not sure ...
I've disabled uPnP in Transmission and it seems to work slightly better but still not good.
I'm reluctant to putting the NAS in DMZ since it becomes to Wide-open then. But I'll test it all the same when I get back home. Not the kind of thing you want to test when you're on vacation. 🙂
Even of this does not help I still want to thank your for your efforts. I really appreciate it.
I 've started to get better speed, 200kb/s plus which is better not close to my max of 1-1.5Mbyte/s which I should be able to get.
So, I checked the daemon.log for more info and since last time I see a lot of "Could not connect to tracker (announcer.c:990)". Afaik this is a problem if the port(s) used are not forwarded in the FW to the NAS (where I run transmission) but I've double checked and "random ports" I've designated in the settings.json file 60000-60999 and the default ports 6880-6999+51413 are forwarded to the NAS so I'm pretty sure this is not the problem. This is with uPnP disabled so I will try again with it enabled.
I find it strange that I do get better DL but still a lot of errors in the logs.
OK, now I'm back at the terrible speeds that I had before and I have no idea why.
I've tried with uPNP enables and disables in Transmission and it does not make any difference. I've tried limiting the Active DL:s to <5, <10, <20, <30, etc, and it doesn't seem to make any difference. I've tried only having the torrents with 1000+ seeds Active, and it doesn't seem to make any difference. I've tried having only smaller (<200Mb) and only larger (4Gb<) torrents Active, but it does not make any difference.
I've naturally stop:ed and started Transmission a number of times to see if it's a port issue, but it does not seem to be so.
In the daemon.log I can still see a lot of the errors above but I can not make heads or tails of if it's worse or less then before. It's simply to many of them.
I gave me a few ideas: 1. Find a good balance of the number of Torrents in Transmission. Still working on that but it seems like Transmission and the ReadyNAS NV+ is at best below 20-30 concurrent Torrents, including the ones you are seeding. 2. Have the ReadeyNAS drop connections faster. This is a bit tricky so does anybody have any suggestions?
I think the problem is found. The main problem is that I had some 400+ torrent Active in Transmission, that is to many. Pausing some 300+ torrent made it better, but not good.
The other problem I think was that I had a UL limit set to approx 80% of my DL, per recommendation to make things faster. I removed this limit and things started to DL at reasonable speed.
However I still was not having a lot of luck connection but when connection I get good DL so then i limited the number of active torrent based on the number of available seed and when I kept the number of active seeds to under approx 2000 seeders for all Active Torrent, things are faster again. Not as fast as expected but faster.
So I'm left to think that I have reached the limitation of the ReadyNAS NV+. Anybody got any input on that?
