microLXC Public Test

191011121315»

Comments

  • somiksomik OG Hostbusters
    edited March 14

    @Neoon said:
    Update on PK:

    I got a reply on my second ticket, this time it wasn't auto closed.
    Apparently its fixed but the VM is still down, I can't access their website right now, so I can't check anything.
    According to them, they lost track of the ticket.

    Anyway, I keep you updated.

    "Sorry, too busy bombing Afghanistan and lost track of your ticket".

    • (probably) the support team :lol:

    I speak fluent sarcasm and broken logic. | I would agree with you, but thæn we’d both be wrong.

  • Restocked: Tokyo, Singapore, Bratislava and Valdivia

  • somiksomik OG Hostbusters

    FYI, alpine linux does not "shutdown" with the "Stop" button in panel, probably cause alpine uses poweroff instead of the standard shutdown -h 0

    I speak fluent sarcasm and broken logic. | I would agree with you, but thæn we’d both be wrong.

  • PK is back online since about 2 days.
    I missed the email from hetrix, however I got a Ticket response yesterday.

    Someone or something fixed PK, but I don't know who, the Support that replied yesterday denies any allegations.
    Basically they just replied that the VM is working and Online, as far as I know last time I checked I couldn't even boot it.

    Anyway, enjoy your PK containers everyone.
    I am unsure at this point if I gonna enable deployments on PK again.

    The Console is currently still broken, I dislike most "fixes", I don't have a propper non hacky solution yet.

  • The console issue has been fixed, sorry that it took so long.
    The patch included security fixes, so I could not hesitate updating incus/lxd.

  • NeoonNeoon OG
    edited April 21

    Tokyo is getting its prefix, tomorraw.

    As stated in previously sent emails, as part of migration of Japan services to our new infrastructure at Equinix TY8 location, the following changes will be applied with regard to IPv6 addresses assigned to existing services:

    • All of the services will automatically receive one dedicated routed /48 IPv6 prefix (once the migration is completed on 22nd April 2026).

    Also

    Maintenance Window:

    Start Time: 1 AM 22nd April 2026 (Japan time)
    Completion Time: 7 AM 22nd April (Japan time)

    Impact: Service will be inaccessible during the migration phase (depending on disk size it can take 5 to 30 minutes per service)

  • somiksomik OG Hostbusters

    @Neoon said:
    • All of the services will automatically receive one dedicated routed /48 IPv6 prefix

    And the SG dedicated server I got from oneprovider just gives a /128 IPv6 and they say they can't even add additional IPv6 IPs to the server...

    Oh well, IPv6 adoption amongst the provider is still very poor...

    I speak fluent sarcasm and broken logic. | I would agree with you, but thæn we’d both be wrong.

  • edited April 21

    @Neoon my Islamabad/PK container cannot be rebooted regardless of me trying from the Dashboard or from the Shell :s can you check? It is still online & accessible though!
    Ah: stopping and starting the container worked!

    💩 VirCrap 💩

  • @jcn50 said:
    @Neoon my Islamabad/PK container cannot be rebooted regardless of me trying from the Dashboard or from the Shell :s can you check? It is still online & accessible though!
    Ah: stopping and starting the container worked!

    You have been rebooting it successfully, according to logs, neither do I have issues rebooting.
    What is the issue?

  • @Neoon said: What is the issue?

    I have a service that starts @ boot time, and whatever I tried it did not start:

    • Dashboard Restart button
    • Dashboard Stop + Start buttons
    • Shell console reboot command

    The uptime command gives the uptime of the host~ not of the container.

    💩 VirCrap 💩

  • Nevermind: everything seems to be working fine now! Sorry for the bother!

    💩 VirCrap 💩

  • @jcn50 said:

    @Neoon said: What is the issue?

    I have a service that starts @ boot time, and whatever I tried it did not start:

    • Dashboard Restart button
    • Dashboard Stop + Start buttons
    • Shell console reboot command

    The uptime command gives the uptime of the host~ not of the container.

    Still, I don't understand what your issue is, your container rebooted fine.
    Yes some tools might give you different info depending on what and what os you are using.

  • I also added PK back in stock, since it has been stable the last 3 weeks.

  • The Subnet in JP is ready and configured.
    I will do a quick reboot tomorrow and enable IPv6 afterwards.

    As soon its available, just press reconfigure in the panel.

  • @Neoon said:
    The Subnet in JP is ready and configured.
    I will do a quick reboot tomorrow and enable IPv6 afterwards.

    As soon its available, just press reconfigure in the panel.

    Done.

  • @Neoon said: The Subnet in JP is ready and configured.

    Thank you, I'm glad it has IPv6 now.

    Here is a minor problem in Tokyo Equinix node however, when I try to install Arch Linux OS to a LXC VPS, the control panel prompted success and the IPv6 is reachable via ping, but SSH failed, and the port seems not open (I tried nmap -6 -p 22 IPV6-ADDRESS).

    In the past, I found sshd in Arch Linux is not enabled via systemd, but since console or shell worked in the control panel, I could manually enable it there, but now console or shell seem not working on my side either. By the way, Debian OS works perfectly. Thank you.

    MicroLXC is lovable.

  • @somik said:
    @Neoon can help to check on the alpine config when you are free?

    Running free -h reports the host's memory in alpine, not the guest alpine OS, while cat /sys/fs/cgroup/memory.max shows the memory correctly. Probably some config issue or some weirdness with Alpine itself...

    lxc5d7a7b88:~# free -h
                  total        used        free      shared  buff/cache   available
    Mem:           7.7G        5.8G      535.7M      361.3M        1.4G      246.5M
    Swap:             0           0           0
    
    lxc5d7a7b88:~# cat /sys/fs/cgroup/memory.max
    268435456
    
    lxc5d7a7b88:~# cat /sys/fs/cgroup/memory.current
    27455488
    

    @jcn50 said: The uptime command gives the uptime of the host~ not of the container.

    This is because Alpine relies on sysinfo for that information.
    To fix it you have to set security.syscalls.intercept.sysinfo=true under the instance setting for each Alpine container and restart it
    reference:
    Instance options - Security policies - security.syscalls.intercept.sysinfo

  • @bliss said:

    @Neoon said: The Subnet in JP is ready and configured.

    Thank you, I'm glad it has IPv6 now.

    Here is a minor problem in Tokyo Equinix node however, when I try to install Arch Linux OS to a LXC VPS, the control panel prompted success and the IPv6 is reachable via ping, but SSH failed, and the port seems not open (I tried nmap -6 -p 22 IPV6-ADDRESS).

    In the past, I found sshd in Arch Linux is not enabled via systemd, but since console or shell worked in the control panel, I could manually enable it there, but now console or shell seem not working on my side either. By the way, Debian OS works perfectly. Thank you.

    its always DNS, the IPv6 of the node changed but nginx cached the DNS response, fixed.

  • @SeederKun said:

    @somik said:
    @Neoon can help to check on the alpine config when you are free?

    Running free -h reports the host's memory in alpine, not the guest alpine OS, while cat /sys/fs/cgroup/memory.max shows the memory correctly. Probably some config issue or some weirdness with Alpine itself...

    lxc5d7a7b88:~# free -h
                  total        used        free      shared  buff/cache   available
    Mem:           7.7G        5.8G      535.7M      361.3M        1.4G      246.5M
    Swap:             0           0           0
    
    lxc5d7a7b88:~# cat /sys/fs/cgroup/memory.max
    268435456
    
    lxc5d7a7b88:~# cat /sys/fs/cgroup/memory.current
    27455488
    

    @jcn50 said: The uptime command gives the uptime of the host~ not of the container.

    This is because Alpine relies on sysinfo for that information.
    To fix it you have to set security.syscalls.intercept.sysinfo=true under the instance setting for each Alpine container and restart it
    reference:
    Instance options - Security policies - security.syscalls.intercept.sysinfo

    Yea, nah, its disabled by default for a reason.
    Its a nice feature to allow a container to display the correct memory usage.
    However it requires a elevated process on the node to respond, which is limited of course but, nah.

  • Lads and Gents, we got a new Node in Ho Chi Minh.
    Courtesy of Onidel, thanks @oloke

    /64 routed iPv6, HAPorxy, DNS Proxy the usual.

  • WSSWSS OG Guru Meditation Error

    @Neoon said:
    Lads and Gents, we got a new Node in Ho Chi Minh.
    Courtesy of Onidel, thanks @oloke

    /64 routed iPv6, HAPorxy, DNS Proxy the usual.

    Richard Nixon back again?

    "It's a hard life- to be a stick insect." - Karl Pilkington

  • somiksomik OG Hostbusters

    @Neoon said:
    Lads and Gents, we got a new Node in Ho Chi Minh.
    Courtesy of Onidel, thanks @oloke

    /64 routed iPv6, HAPorxy, DNS Proxy the usual.

    Is that a China node? Wow!

    I speak fluent sarcasm and broken logic. | I would agree with you, but thæn we’d both be wrong.

  • @somik said:

    @Neoon said:
    Lads and Gents, we got a new Node in Ho Chi Minh.
    Courtesy of Onidel, thanks @oloke

    /64 routed iPv6, HAPorxy, DNS Proxy the usual.

    Is that a China node? Wow!

  • Maintenance Announcement
    We still got 4 nodes running LXD, that are yet not migrated.
    The plan is for the next weekend to migrate them to Incus plus a distro upgrade with a new kernel.

    This includes: Singapore, Tokyo Equinix, Oradea and Islamabad.
    Expect up to 30 minutes downtime per Node, usually less.

    I will start at about 14:00 CET next Sunday.

  • Relocation Announcement
    Our backend currently runs in AMS with Gigahost, Gigahost is gonna discontinue AMS on Wednesday morning.
    Hence our backend gets physically moved to Norway, this should not take more than 24 hours.

    During this time, the Panel including any operations will be unavailable.
    However, containers on the nodes will be running as usual.

    Backups are done, in case something goes wrong on the side of Gigahost, we are good.
    I am gonna take the opportunity, to also containerize and compartmentalize our backend.

    Which means the Panel might be unavailable for a short time today/tomorrow, during the migration.
    Just a heads up.

  • @Neoon said:
    Relocation Announcement
    Our backend currently runs in AMS with Gigahost, Gigahost is gonna discontinue AMS on Wednesday morning.
    Hence our backend gets physically moved to Norway, this should not take more than 24 hours.

    During this time, the Panel including any operations will be unavailable.
    However, containers on the nodes will be running as usual.

    Backups are done, in case something goes wrong on the side of Gigahost, we are good.
    I am gonna take the opportunity, to also containerize and compartmentalize our backend.

    Which means the Panel might be unavailable for a short time today/tomorrow, during the migration.
    Just a heads up.

    Done, its up again.

  • somiksomik OG Hostbusters
    edited April 30

    @Neoon said:
    Done, its up again.

    Erm, ok... so either wait for a while or do it again so it goes down? :lol:

    If it stays up for more then 4 hours, see a doctor!

    I speak fluent sarcasm and broken logic. | I would agree with you, but thæn we’d both be wrong.

  • @Neoon said:
    Maintenance Announcement
    We still got 4 nodes running LXD, that are yet not migrated.
    The plan is for the next weekend to migrate them to Incus plus a distro upgrade with a new kernel.

    This includes: Singapore, Tokyo Equinix, Oradea and Islamabad.
    Expect up to 30 minutes downtime per Node, usually less.

    I will start at about 14:00 CET next Sunday.

    Done.

  • @Neoon said:
    Maintenance Announcement
    We still got 4 nodes running LXD, that are yet not migrated.
    The plan is for the next weekend to migrate them to Incus plus a distro upgrade with a new kernel.

    This includes: Singapore, Tokyo Equinix, Oradea and Islamabad.
    Expect up to 30 minutes downtime per Node, usually less.

    I will start at about 14:00 CET next Sunday.

    I am targeting to upgrade this week Mumbai and Sandefjord, same time.
    Just one reboot will do, about 5 minutes downtime.

  • Just a heads up.

    I set Sandefjord, Helsinki and Meppel for new deployments, to KVM only.
    A Buildserver for KVM templates will happen soonTM, so we can hopefully also offer 128MB and 64MB KVM's.

  • @Neoon said:

    @Neoon said:
    Maintenance Announcement
    We still got 4 nodes running LXD, that are yet not migrated.
    The plan is for the next weekend to migrate them to Incus plus a distro upgrade with a new kernel.

    This includes: Singapore, Tokyo Equinix, Oradea and Islamabad.
    Expect up to 30 minutes downtime per Node, usually less.

    I will start at about 14:00 CET next Sunday.

    I am targeting to upgrade this week Mumbai and Sandefjord, same time.
    Just one reboot will do, about 5 minutes downtime.

    Done.

  • Regarding the failed deployments today and the unavailability of the console / shell.
    A internal cert expired, it has been renewed but this caused deployments to fail and the console / shell to be unavailable for a few hours today.

  • @Neoon said:
    Just a heads up.

    I set Sandefjord, Helsinki and Meppel for new deployments, to KVM only.
    A Buildserver for KVM templates will happen soonTM, so we can hopefully also offer 128MB and 64MB KVM's.

    Kernel module loading for containers has been disabled on nodes where its possible.
    The rest still have to be upgraded to 6.x until we can go and disable it.

    As soon as news hits, these modules will be blacklisted asap anyway by me.
    If you need a certain module loaded, you can send a DM.

    Containers on Sandefjord have been enabled again (with custom packages).
    Meppel will follow once upgraded to 6.x.

    Livepatches for the Kernel are applied as usual, when published.

  • Heads up for the India Node.

    A few days ago, I got a email saying WH is gonna discontinue their old IPv4 subnet.
    I just read this email today, I wasn't send to the correct mailbox, my bad.

    Hence we have to change the IPv4 today, since its gonna stop working tomorrow.
    I wish we had more time but seems to be urgent, its just a week.

    New IP is displayed in the Panel, so you can update your DNS.
    Will be swapped out later today.

  • NeoonNeoon OG
    edited June 18

    WebHorizon: Emergency Maintenance (Singapore)

    We have sent you this email to let you know that we have network card failure on one of our Singapore nodes which has caused servers on that node become inaccessible.

    Maintenance has been enabled for that node and we are investigating the cause to resolve the issue.

  • Heads up for the Romania Node

    There are planned IP address changes, but I don't know when to be precise.
    I will update you when I know more.

  • @Neoon said:
    Heads up for the Romania Node

    There are planned IP address changes, but I don't know when to be precise.
    I will update you when I know more.

    Nevermind, doesn't affect the microLXC node.

  • NL got rebooted and migrated to another Node.
    Live migration wasn't possible due to the cpu passthrough we got enabled.
    Hence the 30 minutes downtime.

    Enjoy the CPU upgrade.

  • Rebooted Helsinki, since the bookworm patch just got available.

  • The Node in India had an outage today due to a fiber cut, both redundant path went down due to unforeseen circumstances, this has been resolved though.

  • Hello @Neoon,
    The HAProxy on Tokyo Equinix node seems not working, I'd appreciate it if you could check out when you are free.
    Thank you.

    MicroLXC is lovable.

    Bump justification: This is an ongoing thread with updates from time to time, now maybe the HAProxy on Tokyo Equinix doesn't work as expected, so I'm asking @Neoon for help.
  • @bliss said:
    Hello @Neoon,
    The HAProxy on Tokyo Equinix node seems not working, I'd appreciate it if you could check out when you are free.
    Thank you.

    I see no problem, please provide a bit more info.

  • @Neoon said:

    @bliss said:
    Hello @Neoon,
    The HAProxy on Tokyo Equinix node seems not working, I'd appreciate it if you could check out when you are free.
    Thank you.

    I see no problem, please provide a bit more info.

    Yes, you are right, it works as expected. I'm sorry for this misreport. I'll check twice in the future, and thank you very much.

    Details are:
    1. I've set up the hostname, pointing to the public IPv4, then opened a test python script on port 80 like this sudo python -m http.server 80, and disabled the firewall for test. But I couldn't visit the hostname with Firefox browser with error code "NS_ERROR_NET_EMPTY_RESPONSE";
    2. Then you replied the service was fine, so I tried with curl -4 http://www.my-domain.com, and it did work fine.

    So I guess I've tuned the config of Firefox too strictly, and I'm sorry for this misunderstanding. It's great the redirect rule on Cloudflare can be removed, and use default 443 instead.

    Thank you and have a nice day, Neoon.

    MicroLXC is lovable.

Sign In or Register to comment.