Nominet Maitenance

[Hay]

Are you peered with nominet? What response times are you seeing?

 

[RobM]

I've tried four different datacenters which all vary in their routes, peering, backbones. They are all slowed by up to 50% - you think it's just me?

 

[dee]

I've tried four different datacenters which all vary in their routes, peering, backbones. They are all slowed by up to 50% - you think it's just me?

You can have 100% only if you deposit 90k

 

[Hay]

Just done some checks and here is what i found:

MTR results:

Packets Pings
Host Loss% Snt Last Avg Best Wrst StDev
1. xx.xx.xxx.xx 0.0% 4 0.3 0.3 0.3 0.4 0.1
2. 195.66.236.61 0.0% 4 0.7 0.8 0.8 1.2 0.8
3. ???

Traceroute / network latency to nominet edge is showing as below 1ms (That is normal for me)

DAC is taking approx 0.003265993 seconds (3.2ms) round trip time according to a packet capture, This is from sending the query to receiving the response to sending the next query

I don't think i am seeing a 50% delay though ive not been keeping track of the DAC round trip times?

 

[RobM]

Yeah the DAC roundtrip time is slower. Ah well not the end of the world just seems a step backwards if they're supposed to be moving forwards.

 

[Hay]

They could have switched over to a backup environment in a different zone or DC last night while they work on their production environment (Just hazard a guess if the RTT has really increased by 50%) either that or they have implemented code that is taking longer to run (Potential logging or additional DB queries)

What round trip times was you seeing prior to yesterday?

 

[RobM]

Less than 3ms - it's fine if it's an increase across the board

 

[Hay]

I'm intrigued to see how this plays out, something fishy going on, i just hope my concerns are not becoming a reality and that nominet are not thinking about imposing a DAC Delay which would be based on account level / finances (like RoR)

 

[Aaron Clifford]

Who knows what their plan is, best not to give them any ideas

 

[RobM]

I'm sure the grubby fingers at the top already have everything mapped out

 

[lazarus]

I am not seeing much of a change. I am asking the DAC a question and receiving an answer between 1.8ms & 2ms for a round trip, which is about right.

An EPP round trip is between 11ms and 14ms

 

[Hay]

An EPP round trip is between 11ms and 14ms

Does that EPP seems quite high to you?

 

[lazarus]

Hi, No, that's about right (I believe). With the EPP you have more bytes to send and receive and slightly more going on at the other end instead of just giving a yes or no answer.

Edit: This EPP timing is a create request whilst already being logged in.

 

[Hay]

Yeah the EPP response is quite heavy

 

[RobM]

Apparently according to nominet everything is fine. So anyone having problems is just imagining it.

 

[lazarus]

Apparently according to nominet everything is fine. So anyone having problems is just imagining it.

Do you think they are playing around with locations with there load balancing?

 

[RobM]

No idea. All I know is I tried four *different* datacenters in varying locations, on three different backbones, and they are all showing slower than they were before the maintenance. That covers several different routes as well. Maybe it's just my tag.... What's clear is that they're not going to do anything although I never really expected them to - there's so many questions they could have asked to try and pinpoint the issue but they didn't so I can see they had no intention of looking further.

 

[philliph]

Has it affected the caching bug/feature? I wondered if the maintenance was to kill it (been out of things for a couple of weeks)

 

[RobM]

Not that I can see. However if a speed slowdown is not occurring for everyone there's cause for concern.

 

[philliph]

Not that I can see. However if a speed slowdown is not occurring for everyone there's cause for concern.

I'm not personally seeing a slowdown (from 2 data centers)