I've got a 3800HGV still kicking from my original install in late 2009. A couple days ago I had a series of short disconnect warnings in a game I play (L4D2, with a 30sec server timeout, so I didn't drop). While playing I'm using Ventrilo, and connected to an IRC channel on a secondary machine. Neither of these disconnected that evening.
Tonight I was playing again and dropped completely. Everything disconnected. I had enough time to login to my router, verify it was up, and attempt to connect to the RG, when the connection came back up. About 10 minutes later it dropped again, for a shorter period this time. Ventrilo's audible connect sound effect makes a great "hey it's back" notification.
The training history in the DSL diagnostics on the RG reveals an exit code of "ERR_HI_BER_LIMIT" around the time of the drop. This is accompanied by accompanied by a large number of CRCs and FECs (I know the first term, but not the second). This seems to be the most concise bit of info I can find in a cursory examination of the unit.
Now obviously these issues don't have to be a problem with the RG, but it got me to thinking that the box will probably fail some day, and I'm wondering how many cattle I'll have to sacrifice to get a new one. Can it be accomplished in less than two hours on the phone with tech support? I'd prefer to replace it *before* it fails completely if that's where this situation is headed.
Aside from that godforsaken box and its determination to forget my DMZ+ setting on occasion, I have to say that this is the first noticeable service degradation I've seen in the entire time I've had the uverse service... I realize this makes me one of the lucky ones.
Thanks for any input you folks might have..
-- edit --
I finally remembered UV realtime (couldn't recall the name) and installed it. Reading through the docs, it appears that my line statistics range from normal to exceedingly good.
The error table shows 2 link retrains with 2 training timeouts, and a huge number of corrected blocks (62272) along with 948 uncorrected blocks. These numbers are stable at this point, so I'm presuming they accumulated in the period where I was having trouble. This data is limited as the RG was rebooted earlier today for unrelated reasons.
Anyone care to guess what happened here? This took place a bit after midnight, so it's not as if they were fiddling with the equipment upstream.
Thanks again.
↧