[atlas] bounces?

Hello,

>> 4520 is a firmware version. It is displayed on the web page for your
>> probe. The controller logged that it finished sending the firmware to
>> the probe at '2013-04-10 01:17:27' and the probe reboots shortly after
>> that to install the firmware. It looks like the probe was up for 43 days
>> before it rebooted.
> 
> Feb 25 14:08:25 sw0.sea.rg.net 3075: Feb 25 14:08:24.566: %LINEPROTO-5-UPDOWN: Line protocol on Interface GigabitEthernet1/0/19, changed state to down
> Feb 25 14:08:25 sw0.sea.rg.net 3076: Feb 25 14:08:25.573: %LINK-3-UPDOWN: Interface GigabitEthernet1/0/19, changed state to down
> Feb 25 14:08:30 sw0.sea.rg.net 3077: Feb 25 14:08:29.633: %LINK-3-UPDOWN: Interface GigabitEthernet1/0/19, changed state to up
> Feb 25 14:08:30 sw0.sea.rg.net 3078: Feb 25 14:08:30.640: %LINEPROTO-5-UPDOWN: Line protocol on Interface GigabitEthernet1/0/19, changed state to up
> Feb 25 14:10:40 sw0.sea.rg.net 3079: Feb 25 14:10:39.130: %LINEPROTO-5-UPDOWN: Line protocol on Interface GigabitEthernet1/0/19, changed state to down
> Feb 25 14:10:42 sw0.sea.rg.net 3080: Feb 25 14:10:41.135: %LINEPROTO-5-UPDOWN: Line protocol on Interface GigabitEthernet1/0/19, changed state to up
> 
> i plead innocent.  i was in singapore at apricot and the probe is in a
> locked cabinet

Probes try to keep up a TCP connection to our infrastructure so that they
can report, and we can involve them in new measurements quickly; as long as
this is up they are shown as "connected", if not, they are shown as
"disconnected".

Such a connection can go down occasionally for a number of reasons:
* the probe has to (decides to) reboot
* the host's network (or home router) has a problem such as a power outage
* the host wants to vaccum up but they don't have enough power sockets free
* we ask the probe to upgrade its firmware
* we do maintenance work on the controller and need to disconnect probes
* the colo where we have infrastructure does network maintenance or has a hickup
* there's a network split between the probe and the controller
* many other reasons

The "downtime" is expected to be a few (up to 20-30) minutes, and these are
the times when the probe may upgrade its firmware. This happens all the
time, it's a part of normal operation, so you don't need to worry if you see
disconnect events.

In the future we plan to switch to measuring the probe "uptime" (or "useful
time") based on the regular measurements it does, rather than pure
connection times.

Regards,
Robert