[atlas] (not) disconnected probes

Hi Romain,

Yep. There is a */4 hour job that tags probes. The job kicks healed 
probes away from the Sanatorium. Then the probes get assigned to a less 
loaded "normal" controller.

You're welcome!
wbr
/vty

On 3/11/20 2:31 PM, Romain Fontugne wrote:
> Hi Viktor,
>
> Thanks, I have learnt something, I didn't know about that Sanatorium 
> controller. Still I am amazed that these probes are doing that at the 
> exact same time, are all the probes tagged "Resolves AAAA/A Correctly" 
> at the same time?
>
> Thanks,
> Romain
>
> On 3/11/20 8:50 PM, Viktor Naumov wrote:
>> Hi Romain,
>>
>> I was wrong, the probe diagnostics works correctly. The probes seem 
>> facing local DNS misconfiguration (cannot resolve controller) or not 
>> allowed to connect to other controllers but Sanatorium. Such probes 
>> connect to Sanatorium controller by IP address. When a probe is 
>> connected to Sanatorium controller and tagged as "Resolves AAAA/A 
>> Correctly" it gets kicked from Sanatorium to a normal controller 
>> which it cannot connect due to mentioned reasons it gets into loop of 
>> disconnects.
>>
>> I also cannot exclude the probe USB flash corruption since it can 
>> lead to weird probe behavior.
>>
>> wbr
>> /vty
>>
>> On 3/11/20 9:32 AM, Viktor Naumov wrote:
>>> Hi Romain,
>>>
>>> I'm checking what the problem can be. Art first glance it looks like 
>>> a false positive in the probe diagnostics system. It causes probe 
>>> redirect to the "Sanatorium" controller and connection flips.
>>>
>>> I will be back to you.
>>>
>>> wbr
>>> /vty
>>>
>>> On 3/11/20 6:04 AM, Romain Fontugne wrote:
>>>> Hi,
>>>>
>>>> I found that the following probes are reported disconnected for 
>>>> several hours a day, more or less at the same time, but all their 
>>>> built-in measurements are there and seem normal:
>>>>
>>>> https://atlas.ripe.net/probes/32639/#!tab-network
>>>> https://atlas.ripe.net/probes/28818/#!tab-network
>>>> https://atlas.ripe.net/probes/19107/#!tab-network
>>>> https://atlas.ripe.net/probes/14831/#!tab-network
>>>> https://atlas.ripe.net/probes/11982/#!tab-network
>>>>
>>>> These probes are in different ISPs and locations in Japan.
>>>> Any idea why this is happening? something funky on the controllers 
>>>> side?
>>>>
>>>> Thanks,
>>>> Romain
>>>>
>>>
>>>
>>
>