0

CLI switching from Ultralow to Anycast

Since yesterday I noticed the CLI on my router (ASUS AX86U with Merlin 386.2) is switching from Ultralow 1 to Anycast 1 after a while (minutes or hours) for some reason.

A restart of the CLI doesn't help but a restart of the router makes it go back to Ultralow, but after a while it's back again to Anycast 1. This have never happened before so something must have changed.

Any idea?! Something changed on the backend?!

■ anexia-cph    4 ms  (anycast2, ultralow1)

  estnoc-cph    5 ms  (ultralow2)

  anexia-osl   33 ms

  zepto-sto    34 ms  (anycast1)

18replies Oldest first
  • Oldest first
  • Newest first
  • Active threads
  • Popular
  • Any theory about this issue?! Still happens every day...

    After a reboot it uses Ultralow-servers as it should:

    ■ estnoc-cph    3 ms  (ultralow2)
    anexia-cph    4 ms  (anycast2, ultralow1)
    anexia-osl   10 ms
    anexia-sto   13 ms
    zepto-sto    16 ms  (anycast1)
    
    "status": "ok",
    "protocol": "DOH",
    "destIP": "193.36.116.10",
    "anycast": false,
    "server": "estnoc-cph-1",
    "clientName": "nextdns-cli",
    

    ...but after a while it switches to Anycast for some reason and not switching back to Ultralow:

    anexia-cph    3 ms  (anycast2, ultralow1)
    estnoc-cph    5 ms  (ultralow2)
    anexia-sto   11 ms
    anexia-osl   12 ms
    ■ zepto-sto    18 ms  (anycast1)
    
    "status": "ok",
    "protocol": "DOH",
    "destIP": "45.90.28.0",
    "anycast": true,
    "server": "zepto-sto-1",
    "clientName": "nextdns-cli",
    

    Any clue what's going on and why it stops using the Ultralow-servers?!

    Like
      • Jörgen
      • Pro subscriber ✅
      • Jorgen_A
      • 2 mths ago
      • Reported - view

      Today suddenly I'm back to Ultralow-servers without a router reboot which was needed before. Is something fixed on the backend or some deployment went wrong?! Will keep an eye on it... 🙂

      Like
    • Jörgen A nothing changed on our side. CLI will fallback on anycast when it can't reach unicast (ultralow). That may happen for different reasons.

      Like 1
      • Jörgen
      • Pro subscriber ✅
      • Jorgen_A
      • 1 mth ago
      • Reported - view

      Olivier Poitrey Thanks for the answer, but still can't understand why the CLI since a few weeks back suddenly prefers the slowest AnyCast-server (400 miles away) over the Ultralow-servers (20 miles away) that it has always used before.

      It finds the fastest servers (as seen below) and after a reboot the CLI uses Ultralow1/2, but after a while it switches over to AnyCast instead. It may not be noticed much in real life, but still annoying knowing I'm not getting the best DNS-servers available. 😉 Is it worth sending a new diag-report for this?!

      •   estnoc-cph    3 ms  (ultralow2)
      •   anexia-cph    4 ms  (anycast2, ultralow1)
      •   anexia-osl   12 ms
      •   anexia-sto   14 ms
      •   zepto-osl    17 ms
      •   zepto-ams    18 ms
      • ■ zepto-sto    24 ms  (anycast1)
      Like
    • Jörgen A I have been noticing this too for the past few days. Fortunately it fails over to anycast 1 which is generally as quick as the low latency servers. Anycast2 is over 100ms so I’d be in trouble if it was picking that. Nothing in the logs indicates why it changes, I just get the usual changing endpoint message. It seems to have started after I updated to v1.32 of the CLI. Which version are you running?

      Like 1
      • Jörgen
      • Pro subscriber ✅
      • Jorgen_A
      • 1 mth ago
      • Reported - view

      Chris Leidich I'm also on 1.32 since it was released (and 1.12.x before that), so could be a clue to Olivier Poitrey . Nothing in my logs either, but something must have changed either in the CLI or the routing/backend.

      Like
    • Jörgen A it’s due to a bug in the HTTP/3 client. See https://github.com/lucas-clemente/quic-go/pull/3157. A new version will come soon with DoH3 disabled, waiting for the fix to be accepted.

      Like 2
    • Olivier Poitrey HTTP/3 is going to fight you every step of the way isn’t it? :) Thanks for the info!

      Like 1
      • Jörgen
      • Pro subscriber ✅
      • Jorgen_A
      • 1 mth ago
      • 1
      • Reported - view

      Olivier Poitrey I installed the new 1.32.1 version yesterday, but when checking this morning it's back to the slowest AnyCast-server like before. Strange... Thanks for you work trying to resolve this! ☺️

      Like 1
      • Maghuro
      • maghuro
      • 1 mth ago
      • 1
      • Reported - view

      Olivier Poitrey same here, even with doh3 disabled, it keeps changing from UL to anycast quadrupling my pings to 80ms

      Like 1
    • Jörgen I am seeing the same behavior on a UDM after upgrade to 1.32.1 - not causing an issue, per say, but it definitely doesn’t stay on Ultralow for more than a few hours after I restart the CLI. 

      Like 1
      • Jörgen
      • Pro subscriber ✅
      • Jorgen_A
      • 4 wk ago
      • Reported - view

      NextDNS Any news or identified the issue why everyone is falling back to Anycast-servers nowadays and not using the faster Ultralow-servers?

      Like
  • nextdns-cli (1.32.1) on pfsense 2.5.1 switches every time from ultra low to anycast exact 2 hours after starting nextdns-cli. 

    Like 1
      • Maghuro
      • maghuro
      • 1 mth ago
      • 1
      • Reported - view

      Wouter De Visschere same on merlin

      Like 1
    • exact every to hours a switch between anycast <-> ultra low

      Like
  • Same problem yet. Any news?

    Like
  • Can you try disabling CNAME Flattening?  I'm getting fastest server.  Will further test to make sure

    Like
  • Bumping this old thread, since the issue should be fixed now. Thanks NextDNS team!! 👏

    If not working right away, give the router a reboot.

    Like
Like Follow
  • 13 days agoLast active
  • 18Replies
  • 413Views
  • 6 Following