Hello admins of EU AWS systems – DNS outage in eu-west-1

danielhunt · 2016-12-15T02:23:32.000Z

https://status.aws.amazon.comLooks like there's a DNS outage in AWS DublinAll DNS request coming from `eu-west-1b` fail completely if you're using the default AWS DNS resolution itself - if you switch to `8.8.8.8` it's all totally fine, but that's of no use if you have private services that need internal AWS resolution`1a` and `1c` are also affected, but to a much lower extent from what I can tell

aidos 9 years ago |

Question for the more experienced in the group - is there a way of caching the DNS locally within you VPC in case something like this happens again? Or does that just cause more issues than it solves?

When I discovered it I tried switching my resolv.conf to 8.8.8.8 but of course none of my internal stuff worked because of how my security group / subnet / ip restrictions are setup internally.

jlgaddis 9 years ago | |

I'm not a big user of AWS but I don't know that there's anything stopping you from running your own recursive resolver (in your VPC) that your other machines use (instead of using an external resolver or whatever Amazon provides).

You could, for instance, stand up an instance or two running Unbound, forward to Amazon's own resolvers for internal stuff, and forward to external resolvers for external resolution (or just handle it yourself instead of using forwarders).

tomekit 9 years ago | | |

The main reason to use AWS is so you don't need to do anything except calculate profit from your app, at least in theory. If you need to set up own resolvers or anything else within the AWS there is no reason to use AWS in the first place.

chatmasta 9 years ago | |

Yes, you can specify DNS server to use in the DHCP options set of the VPC. I've done this to point to internal DNS servers.

fern4lvarez 9 years ago |

It's weird, I see actually most of my issues coming from instances located in `eu-west-1a`, whereas `eu-west-1b` and `eu-west-1c` look fine.

aidos 9 years ago | |

I recall reading somewhere many years ago that AWS randomise a, b, c etc per account. Otherwise people have a tendency to pick A and be done with it.

danielhunt 9 years ago | | |

That's true - they loadbalance instance creation unless you specifically ask for an instance in a zone

That doesn't apply here though. Instances that are already up are experiencing DNS resolution issues because the Amazon-provided DNS service (which is distinct from Route53) is failing

aidos 9 years ago | | |

I don't think I explained that very clearly.

What you see as zone A is not what I see as zone A (maybe). When you sign up for an account, AWS assigns your zone A to one of the 3 available zones:

    zone X (real) : your A (virtual) : my C (virtual)
    zone Y (real) : your B (virtual) : my B (virtual)
    zone Z (real) : your C (virtual) : my A (virtual)

That's why people see different outage characteristics between the zones.

Edit: one of the few articles (other than HN comments) that explains it https://alestic.com/2009/07/ec2-availability-zones/

danielhunt 9 years ago | | |

I've been using AWS for years and this is the first I've ever heard of this.

All I can say is "WHAAAAAAT?!"

You've made me question my reality. Take an upvote for that alone.

misframer 9 years ago | | |

It's in the documentation [0].

    An Availability Zone is represented by a region code followed by a letter identifier;
    for example, us-east-1a. To ensure that resources are distributed across the Availability
    Zones for a region, we independently map Availability Zones to identifiers for each
    account. For example, your Availability Zone us-east-1a might not be the same location
    as us-east-1a for another account. There's no way for you to coordinate Availability
    Zones between accounts.

[0] http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-reg...

aidos 9 years ago | | |

I'm glad I could make what is otherwise a bad time to be looking after servers a little more enjoyable.

I myself was in the middle of a release when all this happened. "I'll just release this thing" I said, "will only take 20 minutes" I said....

danielhunt 9 years ago | |

Interesting difference in experience there

Complicated networks are complicated, I suppose

danielhunt 9 years ago | |

Seeing an increase in errors in `1a` and `1c` now myself too

neo2001 9 years ago |

6:29 PM PST We have identified the root cause of the DNS resolution issues in the EU-WEST-1 Region and continue working towards resolution.

paugay 9 years ago |

same here :) let's see how long it takes

neo2001 9 years ago |

sigh