quick usage comparison....

You say "and also some more local logging to USB" ... Is this an indication that throughout your random restarts, you have a USB stick attached?

I too have logging to USB (and to a syslog server on a PI). Neither have ever given me a clue to the restart causes. I also use the USB for config backups.

However - I've always have a USB stick attached, and therefore always when I've had reboots. If the same is true for you, could the use of USB be connected?

I use original Sandisk thumbnail usb memory...(so it shouldn't be media quality causing the issue).

It's not related to USB logging as it has occurred both with & without a USB drive connected.

I don't think the config backups are also going to the USB, I'll check, as these go to ACS.

The USB syslog also does not show anything, probably due to the log write being cached and not immediately being written to disk. This indicated something that happens quickly rather than building up quickly like a memory leak.

The only thing I've not been logging is the DSL connection itself, so will look to see what this is doing in more detail immediately prior to a crash.

I only have an issue with one 2862 out of many installed so either a device issue or the site environment.

Will also look at swapping the power supply.

My 2926 reboots itself randomly every couple of days. I have no USB devices attached. (penned in response to e.ringrose above who says he/she has no problems with a 2926).

I have two 2955 boxes at two different sites. Both reboot randomly a few times a day. Went on for years. Swapping the router (I have a spare for each) makes no difference. One of the sites was moved from ADSL over copper to FTTP and that reduced the reboot to about once a day.

I concluded that, with everything on the internet getting hacked a few times each second, it was some rogue packet which the box didn't know what to do with. But why better on FTTP?

Both had quite a different config, and different ISPs.

Both had different ADSL modems. One was Zyxel and the other Vigor 130.

One site has just been changed to a 2960 which messed itself up after a few days and ran extremely slowly, slowly enough for even pings to not work 99% of the time. I had to drive there last night and restart it. The config menu was running fine.

If this carries on I might schedule a reboot every day at 5am or something like that. I also have a power switching box which enables me to interrupt the mains for 5 secs, with an SMS message. Not used it for years but I will get it going again. Not really acceptable for a £400 box.

As various have said, of course it's not acceptable - i.e. you shouldn't accept it, but it can also be immensely difficult to isolate and identify the cause of what is not a common problem. I'd suggest that people do keep up the requests to support, asking that it be escalated, investigated etc. Don't be a pain of course, but do periodically check on progress and don't give up... though of course, that's your prerogative - you're not a paid tester, but we do all benefit from your efforts!

I had this problem at one point, which was caused by USB stick, but that was fixed in newer firmware (DrayOS platform) but I can see that's not always the cause. Eliminating different config. settings one by one as a control can help diagnose, starting from a very plain config.

Well... the standard answer from tech support is: do a factory reset and re-enter the config manually. Do NOT load the config file.

That unfortunately is a total waste of a life (at least half a day) and a really bad way to treat a customer. It is a standard "call centre script monkey" type of approach and one would expect better from Draytek. It also assumes you have documented the config with screenshots.

More subtly it is also an admission that the box has a raft of bugs in the way different features interact and perhaps the order of the configuration (e.g. setting up the ADSL before setting up VPNs) matters.

And I have never ever found it cures anything. Half my working day is spent on this kind of thing.

I was very happy with the 2955 (a quiet reboot didn't affect much other than break long file transfers which we rarely do over a VPN and on the Draytek OS it took only about 10 secs; more like 5 mins on the Linux boxes like the 2960) and it was changed out only because of the very subtle bug which messed up VOIP by dropping UDP for 10 secs at a time. I spent a week or two of my life on that.

There is no easy solution, short of spending 10x as much on an enterprise level box from Cisco. The sub-1k boxes from Cisco are rebadged junk and full of bugs. Read the reviews on CISCO RV320 for a taster; luckily we wasted just one day on that one before returning it.