DNS servers failing

[#1]

Quote from: spoon

Then please explain why people who were using OpenDNS or Google DNS were also affected.

Did they refresh their DNS caches?

[#2]

Quote from: AndyH

I have two separate Plusnet connections - both were unaffected by the 'issue', but both reported around 38-42% DNS failures in Samknows between 01:00 and 08:00 (although every other test was fine).

My Samknows shows no DNS failures in the past week (currently using 212.159.6.10 and 212.159.6.9). Confused.

[#3]

Quote from: AndyH

I can't see what it would have to do with BT's infrastructure - it's Plusnet's DNS servers that were the issue.

Was it?

Quote from: http

Our engineers are investigating alongside our suppliers, and once we have more information, we'll provide a further update.

Quote from: AndyH

I have two separate Plusnet connections - both were unaffected by the 'issue',

Both rather hint at the possibility of it being without PlusNet.
The point is that none of us know and PlusNet are not telling and we are ALL just speculating.  Would be so much healthier if someone said something to dispel the speculation... at least it might enable the smart ones around here to watch out for tell-tale signs which might assist in diagnosing issues.

[#4]

Quote

My Samknows shows no DNS failures in the past week

Mine does, between midnight & 08.30.

[#5]

Note Failed DNS queries doesn't mean the DNS itself failed, it could be that the DNS server couldn't be reached.

[#6]

I think not everyone was affected by the outage, and not everyone who was affected, was affected in the same way. At about 6 or 7am, I found most but not all websites were not loading, and switching to Google DNS or OpenDNS did fix the issue for me. Plusnet may have partly fixed things by that time.
During the outage, I saw that Plusnet's DNS servers were returning SERVFAIL responses. I think that indicates I could reach the Plusnet DNS servers, but the Plusnet DNS servers could not reach the external nameservers they needed to complete the lookup.

[#7]

That makes sense and would tie in with people using other DNS servers being unable to reach some sites - i.e. some sort of routing issue.

[#8]

Quote from: ejs

I think not everyone was affected by the outage

That would be my conclusion since I was totally unaware of any outage until I saw it on here this morning (I've been online using Plusnet DNS - no issues -  but not on the forums for that past day or 2).

[#9]

Quote from: ejs

I think not everyone was affected by the outage,

On the contrary, it may well be that they were ALL affected but most sleep during normal sleeping hours - unlike those of us who were really affected by it suddenly dropping dead 

[#10]

It's likely only one site was affected - I'm connected to a pcl gateway and the SamKnows box logged DNS failures. Perhaps if people know which gateway they were on during the failure they could post and say if they were affected.
(Of course if this were the case PN could put an end to such speculation by posting to say so!)

[#11]

Quote from: 7up

On the contrary, it may well be that they were ALL affected but most sleep during normal sleeping hours - unlike those of us who were really affected by it suddenly dropping dead

Not correct otherwise all the BQM graphs would have shown it

[#12]

Quote from: 7up

On the contrary, it may well be that they were ALL affected but most sleep during normal sleeping hours

Nah, she quite usually sits up 'til 'stupid-o'clock 'Facebooking' on her phone, don't think that night was any different.

Quote from: Oldjim

Not correct otherwise all the BQM graphs would have shown it

Absolutely, my BQM is perfect for the evening of the 9th / morning of the 10th (I assume I have the right dates) below is 10th Sept:

[#13]

Quote

Perhaps if people know which gateway they were on during the failure they could post and say if they were affected.

I was on Ptn-ag02 & was affected. It was OK when I changed to Google, although much slower than ususal.

[#14]

I've been on pcl-ag02 since 07/09.

[#15]

Quote from: Oldjim

Not correct otherwise all the BQM graphs would have shown it

Jim,
That would depend on what the problem was.  I saw the problem but had no issue connecting to thinkbroadband.com - it is that experience that leads me to believe this was a routing issue and not a DNS issue.
I suggest that depending on which route out of PlusNet to the DNS supplied IP address would be used determined if connectivity would be successful.
I noted this morning that the addresses given for one of the popular targets are completely different between PlusNet's DNS and Google's DNS - they appear to be in separate sub-nets with no overlap.  Therefore I suggest it is possible that PlusNet could route to one subnet and not the other.
This is all speculation - shame no one can (or is that will?) bring facts to the discussion.
Kevin