nanog mailing list archives

Re: Remote power cycle recommendations


From: Stephen Satchell <list () satchell net>
Date: Mon, 30 Apr 2018 11:58:13 -0700

On 04/30/2018 10:05 AM, William Herrin wrote:
On Mon, Apr 30, 2018 at 12:19 PM, Brielle Bruns<bruns () 2mbit com>  wrote:
In particular, if at all possible, do not use the AP9606 era cards with the
APCs.  They are 10BaseT and take fragile to a whole new level.  I usually
have to manually force the port to 10 on the switch or put it on a cheap
dumb older switch.
They're fragile but they're not_that_  fragile. A switch that can't
figure out 10 mbps half duplex... now that's fragile.

Personally, I've not run into THAT problem in years. What I have run into is when you have a 10base-T target and you connect it to a 100base-T (or faster) infrastructure, the switch as part of the rate changing will tend to flood the poor embedded stack if your application layer isn't very, very careful to space out packets.

At best your embedded-stack device will lose packets. At worst, you will have to power-cycle the poor dear in order to get it to start listening to the network again.

Let me repeat, this observation is not restricted to the AP9606 cards; it seems to be an issue with embedded-stack devices in general.

Subject change: one other thing about the AP9606 cards: they have a battery on them, and you do have to change that battery every decade or so...


Current thread: