[atlas] accounting glitch for probe 15105

Hi Kasper,

You unfortunately hit a special case.
What happened was that your probe was connected and disconnected at the 
same exactly second [1].
This resulted in having two datetime fields in our database with the 
same value for your probe, which in turn made your probe considered as 
disconnected.
Unfortunately the fact that these fields had the same value prevented 
the mechanism that tries to fix this kind of cases from working.
When you reconnected your probe everything went back to normal, and 
probe's uptime history was corrected.
The only issue that was not resolved by itself was crediting as you 
mentioned.
But if you are in need of more credits we would be happy to  give you a 
bunch as a compensation.

Of course, we will work on making sure these special cases will not 
happen in the future.

Regards,
Andreas Strikos
RIPE NCC

[1] It happens some times that signal that ssh connection went down 
comes a bit later than the actual time, which results in cases like these.

On 25/06/16 01:26, Kasper Pedersen wrote:
> Tonight Freddy Bulow's post made me take a look at 15105. I was certain it would be up, but that was only halfway what I found.
>
> It was listed as being down since June 16th, and the uptime graph was showing an 8-day hole. It was, however, showing traffic on the graph, and locally I was also seeing normal probe traffic. Here's a pcap:
> https://highlyexperimental.kitsune.dk/201606242131.pcap
>
> At packet 765 I powercycled the probe, it came back up, and now showed 0 minutes up. That was expected, but what also happened was that the uptime graph filled in to no longer show downtime, the log entry showing the probe going down on the 16th disappeared (yes, really), and a new one showed it going down on the 24th with a few seconds downtime.
>
> The only thing left is the 8-day hole in the credit log, that did not change.
>
> /Kasper Pedersen
> Beware of corner cases. They tend to have sharp corners.
>
> Sent via RIPE Forum -- https://www.ripe.net/participate/mail/forum
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: </ripe/mail/archives/ripe-atlas/attachments/20160628/15ecec36/attachment.html>