Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Sign In Register

Quick Links


Categories

In this Discussion

› › Ubuntu 18.04 network speed issue with Linode and DigitalOcean. Help needed.
New on LowEndTalk? Please Register and read our Community Rules.

Ubuntu 18.04 network speed issue with Linode and DigitalOcean. Help needed.

nqservicesnqservices Member
edited October 2018 in Help

Hi all

I’m have notice some issues with network speed using Ubuntu 18.04 on a Linode VPS. After some tests it seems the issue is with Ubuntu 18.04 because on DigitalOcean almost the same happens (there with Ubuntu 18.04 I also get slow network speed).

So, I would really appreciate if anyone here on LET with a Linode account could do the bellow simple test to try to reproduce the issue.

1- Create a new Linode with Ubuntu 18.04 on London DC and right after login and before any updates run both bellow speed tests:

wget freevps.us/downloads/bench.sh -O - -o /dev/null|bash

wget --no-check-certificate bench.sh && mv "index.html" "bench.sh" && chmod +x bench.sh && ./bench.sh

2- After tests are done run the command:
apt-get update && apt-get upgrade

3- Reboot

4- After reboot make the exact same speed tests you made on step 1. Check and compare if your speed is now very slow (except from Cachefly CDN that I suppose in on the same DC as Linode).

Thanks

Thanked by 1gol3m

Comments

  • MikePTMikePT Moderator, Patron Provider

    That seems to be an issue with the kernel. I have seen similar issues and its not exactly related with network/peering. Run your tests in rescue / using another image if needed. I am afraid only the kernel devs can fix it :-).

  • nqservicesnqservices Member

    @MikePT said:
    That seems to be an issue with the kernel. I have seen similar issues and its not exactly related with network/peering. Run your tests in rescue / using another image if needed. I am afraid only the kernel devs can fix it :-).

    My idea is similar to your's. It seems something related with the OS/Kernel and not network. Do you have any Linode acocunt where you can test this? I really need someone else to confirm this issue.

  • MikePTMikePT Moderator, Patron Provider

    @nqservices said:

    @MikePT said:
    That seems to be an issue with the kernel. I have seen similar issues and its not exactly related with network/peering. Run your tests in rescue / using another image if needed. I am afraid only the kernel devs can fix it :-).

    My idea is similar to your's. It seems something related with the OS/Kernel and not network. Do you have any Linode acocunt where you can test this? I really need someone else to confirm this issue.

    Just launch another vm and test it. Or report to Linode, they will probably investigate it too.

  • nqservicesnqservices Member

    @MikePT said:

    Just launch another vm and test it. Or report to Linode, they will probably investigate it too.

    Made that. Doing that...

  • teamaccteamacc Member

    Have issues with 4.15.0-36 although in -34 it worked fine.

    On any host (tried 10 different hosts). The -lowlatency kernel doesnt fix shit.

  • nqservicesnqservices Member

    Made now 2 more tests using Ubuntu 16.04 on both Linode and DigitalOcean. The network speed on the tests was great! Also made the test using Debian and all seems great.

    Issue only seems to happen wit Ubuntu 18.04.

  • dragon1993dragon1993 Member

    @teamacc said:
    Have issues with 4.15.0-36 although in -34 it worked fine.

    On any host (tried 10 different hosts). The -lowlatency kernel doesnt fix shit.

    Me too

  • connercgconnercg Member
    edited October 2018

    Digital Ocean @ NYC1 fresh installs of both ..

    OS              : Ubuntu 18.04.1 LTS
Arch            : x86_64 (64 Bit)
Kernel          : 4.15.0-36-generic
Hostname        : ubuntu-s-1vcpu-1gb-nyc1-01


Speedtest (IPv4 only)
---------------------
Your public IPv4 is 204.48.21.xxx

Location                Provider        Speed
CDN                     Cachefly        14.5MB/s


Atlanta, GA, US         Coloat          1.52MB/s 
Dallas, TX, US          Softlayer       4.01MB/s 
Seattle, WA, US         Softlayer       2.74MB/s 
San Jose, CA, US        Softlayer       2.50MB/s 
Washington, DC, US      Softlayer       6.81MB/s 

    OS              : Ubuntu 16.04.5 LTS
Arch            : x86_64 (64 Bit)
Kernel          : 4.4.0-137-generic
Hostname        : ubuntu-s-1vcpu-1gb-nyc1-01


Speedtest (IPv4 only)
---------------------
Your public IPv4 is 142.93.246.xxx

Location                Provider        Speed
CDN                     Cachefly        90.6MB/s

Atlanta, GA, US         Coloat          18.0MB/s 
Dallas, TX, US          Softlayer       32.4MB/s 
Seattle, WA, US         Softlayer       27.8MB/s 
San Jose, CA, US        Softlayer       20.1MB/s 
Washington, DC, US      Softlayer       74.9MB/s
    Thanked by 1nqservices
  • MikePTMikePT Moderator, Patron Provider
    edited October 2018

    Definitely the kernel. Should be reported to Ubuntu so they can sort it.

    Should already be reported in the kernel community tho. No idea.

  • nqservicesnqservices Member
    edited October 2018

    Thank you @connercg for the tests! They confirm the issue! If you have the time, please report that to DigitalOcean team so there is more people reporting this for them to give this more priority.

    As @MikePT says it may be related with kernel. I have already reported this both to Linode and DigitalOcean support teams in the last hour.

    Besides this situation sucks, there is a lot os people running their Ubuntu 18.04 VPS's with slow network speed... not only on Linode and DigitalOcean...

    Thanked by 2MikePT ferri
  • JoeMeritJoeMerit Member

    latest wget is messed up in Ubuntu 18.04.... other network tests are just as good as any other version of Ubuntu... i.e. speedtest-cli

  • jarjar Member, Patron Provider
    edited October 2018

    nqservices said: If you have the time, please report that to DigitalOcean team so there is more people reporting this for them to give this more priority.

    That won't increase priority, extra reports will not be needed. We use Ubuntu's cloud server image, so it's unlikely that the impact will be made on our end as it's more likely an issue to be resolved by the distro maintainers. That said, I'll certainly ask Ryan about it and see what he thinks.

    Thanked by 2nqservices MikePT
  • nqservicesnqservices Member
    edited October 2018

    @jar said:
    That said, I'll certainly ask Ryan about it and see what he thinks.

    Thanks!

  • HostEONSHostEONS Member, Host Rep
    edited October 2018

    can you check ethtool eth0 and see what link speed is it detecting and also if Auto-negotiation is on or off

  • nqservicesnqservices Member

    @HostEONS said:
    can you check ethtool eth0 and see what link speed is it detecting and also if Auto-negotiation is on or off

    [email protected]:~# ethtool eth0
    Settings for eth0:
    Supported ports: [ ]
    Supported link modes: Not reported
    Supported pause frame use: No
    Supports auto-negotiation: No
    Supported FEC modes: Not reported
    Advertised link modes: Not reported
    Advertised pause frame use: No
    Advertised auto-negotiation: No
    Advertised FEC modes: Not reported
    Speed: Unknown!
    Duplex: Unknown! (255)
    Port: Other
    PHYAD: 0
    Transceiver: internal
    Auto-negotiation: off
    Link detected: yes

  • HostEONSHostEONS Member, Host Rep

    @nqservices said:

    @HostEONS said:
    can you check ethtool eth0 and see what link speed is it detecting and also if Auto-negotiation is on or off

    [email protected]:~# ethtool eth0
    Settings for eth0:
    Supported ports: [ ]
    Supported link modes: Not reported
    Supported pause frame use: No
    Supports auto-negotiation: No
    Supported FEC modes: Not reported
    Advertised link modes: Not reported
    Advertised pause frame use: No
    Advertised auto-negotiation: No
    Advertised FEC modes: Not reported
    Speed: Unknown!
    Duplex: Unknown! (255)
    Port: Other
    PHYAD: 0
    Transceiver: internal
    Auto-negotiation: off
    Link detected: yes

    I've never tried on a VM, but try following it may or may not fix it:

    ethtool -s eth0 autoneg on

    Thanks

  • HostEONSHostEONS Member, Host Rep

    Also jus make sure the NIC name is eth0 and not something else

  • gol3mgol3m Member

    I had that issue with Contabo and Hetzner about 2 weeks ago. Thanks to you I finally know what caused it.

  • rm_rm_ Member, IPv6 Advocate
    edited October 2018

    It might be this: https://patchwork.ozlabs.org/patch/846763/

    I got the issue with network speeds on kernel 4.14.50: getting at most ~30 mbit incoming. After upgrading to 4.14.51 which contains the above patch, it went away.

    Sure Ubuntu uses 4.15, but likely its bugs and fixes are the same or similar.

  • nqservicesnqservices Member

    @dragon1993 said:

    @teamacc said:
    Have issues with 4.15.0-36 although in -34 it worked fine.

    On any host (tried 10 different hosts). The -lowlatency kernel doesnt fix shit.

    Me too

    The kernel version on Linode is: 4.15.0-36-generic

    @HostEONS said:
    I've never tried on a VM, but try following it may or may not fix it:

    ethtool -s eth0 autoneg on

    Not working:
    [email protected]:~# ethtool -s eth0 autoneg on
    Cannot set new settings: Invalid argument
    not setting autoneg

  • nqservicesnqservices Member
    edited October 2018

    After talking with Linode support team they have confirmed the issue with some of the kernels. As for example the issue happens with kernel "4.15.0-36-generic", but when you change to the latest linode kernel 4.18.8 the speed is great!

  • dragon1993dragon1993 Member

    https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1796895

    Thanked by 1angstrom
  • angstromangstrom Moderator

    Until the next kernel update (which will hopefully fix the regression), it's always possible to downgrade to the previous kernel update, which is normally still present in /boot.

  • HarambeHarambe Member

    This issue also impacted the Proxmox kernel, believe it was the 4.15.18-5-pve kernel where you just had network speeds crap out - at least on the Super Micro boards I'm using.

    Rolling back to -4 or upgrading to the -7 release fixed the issue. Was a frustrating hour or two trying to figure out why my network speeds ate crap randomly.

  • gol3mgol3m Member

    Yesterday's Ubuntu 18.04 update to 4.15.0-38 fixed it.

    Thanked by 1eva2000
Sign In or Register to comment.