I have a chain of devices connected to the Internet:
- Draytek 2120n-plus Fibre router in a Metal Shed - Internet connection
- TP-Link AV500 Powerline Adapter
- Draytek 2760 VDSL Router, used as Wireless Hub
- Linksys SPA2102 VOIP router
These devices are linked up without any special settings that I can remember. The VOIP router is configured according to my VOIP providers guide: http://blog.2talk.co.nz/spa2102.html
Generally, when computers are connected to 1. or 3. through Wifi they work very well. However on 3. I found that browsing, Skype, SSH sessions and any other application regularly "stall". No data seems to be flowing for about 30-90 seconds after which everything is back to normal. This happens every 10 minutes or so.
To troubleshoot this I reduced the MTU on 3. to 1300. This seems to have fixed the stalling problem, however I can now no longer use the VOIP device, I get no dial tone, nor any ringing out.
An Internet search indicates that there is no MTU setting in the SPA 2102 router.
I could also not find any MTU size for Ethernet over Powerline, so I presume this runs on an MTU of 1500?
How would I determine what an appropriate MTU is, I guess it should be 1492, but obviously I also want to eliminate the stalling
Some information on further network devices:
- has a simple switch connected with 1-2 computers and a printer, however the switch is not in between 1. and 2.
- has usually 2 mobile phones connected by Wifi and sometimes another 2 computers.
I do not think this is an MTU problem. MTU problems are usually not intermittent, as the problem you describe seems to be. MTU problems are noticed when existing sessions stop working at usually the same point, such as when getting an image. Wireshark will show that DNS mostly works and that TCP sessions (HTTPS for example) get connected and then start timing out -- but every single time, not every ten minutes.
To test for an MTU problem, you can use a simple
ping HOST -f -l SIZE
, described here in detail by TP-Link. Note that the number used by the Windows ping is the payload, and that headers are 28 bytes, so an MTU of 1500 corresponds to a successfulping -f -l 1472
. Testing should not only be toward the tp-link host as suggested, but between a computer connected to your device 1 and another connected to your device 3. Your internal pings should be OK at 1472, and maybe even to the Internet, but if it is less toward the Internet it is not a problem.Once you have your pings set up (I suggest 1-to-Internet and 1-to-3), keep them pinging each second (add
-t
to the command line), and wait for the interruption to occur to see what happens.So if it's not MTU, what can it be? I suspect either a rogue device that is taking over one of your internal IPs, or a problem with the PowerLine adapter, maybe due to some electrical device that causes a problem every ten minutes or so, and which forces the adapters to re-synchronize. My crystal ball is not perfect, though!
I'm not really answering your question, but this was too long for a comment, and hopefully it will help you find your problem.
This turned out to be a problem of Wifi congestion.
Firstly, with help of Draytek support, I narrowed down the problem to the device and connection method:
As the next step we tried different Wifi channels on Device 3 wifi, specifically CHannel 1 and 11 have less overlaps with other channels. This improved the problem to some degree, but not eliminated it.
I also analysed the signal strength of surrounding Wifi access points with an Android tool: Wifi Analyzer
This confirmed that there are a lot of Wifi access points around and that Channels 1 and 11 were somewhat better.
To resolve these issues I found a 802.11ac dual band router I had lying around (2.4Ghz and 5Ghz) and use this now as a new access point. I also had to buy a 5Ghz Wifi usb dongle for my laptop.