CentOS 8 on OVH - No network connection after cPanel installation

[clopezi]

Hi,

We have a brand new dedicated server on OVH, and we install it with CentOS 8, because it's the only RHEL branch that it's supported there, so I we want to install CloudLinux 8 or AlmaLinux, this is the only way.

CentOS 8 on OVH it's installed with Cloud-init DHCP configuration, and this is causing that after cPanel install, external connection it's lost (but KVM access to the server it's fine, and boot ok)

We have figured out to remove the network configuration from Cloud-init to establish static IP with instance-data.json file. We have rebooted the server some times, and works fine. But after cPanel install... network it's lost again.

OVH it's one of the biggest server providers in the world, so it's very rare that this problem was not happening before and for sure it's gonna happen more often in a near future, so maybe cPanel want to talk with OVH or support this option in some way.

If we install the machine with cPanel template (CentOS 7) or install CentOS 7 and setup it with cPanel, everything it's working fine, without any kind of problems.

Yesterday I opened a ticket (ID 94339458), and 5 people has answered it after many many hours of initial wait. Only one person (thanks Daniel) has tried to resolve it, the others only says "it's your problem or OVH problem" but no more, and OVH says that the installation it's fine. I remember some years ago when the cPanel support was very very good, now I'm sorry to say that, but the price is higher and the support it's not very friendly, slow and not very helpful, right opposite to CloudLinux support.

Do you have any advise? Thanks!

 

[andrew.n]

Although not with OVH but I had similar issues when I removed network manager after installing CentOS 8 and converting it over to AlmaLinux thereafter.

@cPRex might want to have a closer look at the ticket cos I'm not sure it would be a good idea to ask for the content of network configs here.

 

[clopezi]

We have figured out, so I want to share the solution just in case someone found it useful:

1) Edit /run/cloud-init/instance-data.json an remove the IPv4 DHCP settings
2) Run ip addr to see what interfaces aire in use for IPv4 and IPv6, in our case was enp5s0f0
3) Remove old interfaces without use like eth0
4) Edit /etc/sysconfig/network-scripts/ifcfg-enp5s0f0 and set something like this:

DEFROUTE=yes
DEVICE=enp5s0f0
HWADDR=a0:42:3f:...
IPV6ADDR=2001:41d0...
IPV6INIT=yes
IPV6_AUTOCONF=no
IPV6_DEFAULTGW=2001:41d0:...
IPV6_FORCE_ACCEPT_RA=no
ONBOOT=yes
TYPE=Ethernet
USERCTL=no
BOOTPROTO=static
IPADDR=54.xxx
NETMASK=255.255.255.0
GATEWAY=54.xxx.254 (in OVH, same like IPv4 with 254 at end)
MTU=1500
STARTMODE=auto
IPV4_FAILURE_FATAL=yes

We still have some errors, like "RTNETLINK answers: File exists" on systemctl status network, but it's working.

However, cPanel should support OVH configuration out of the box, I'm sure that a important number of cPanel users are on OVH...

 

[andrew.n]

Very well.

 

[cPRex]

@andrew.n - we're definitely aware of issues people run into when installing CentOS 7 and CentOS 8. The problem for us is that each hosting provider, and in some cases even products under the same host (such as Amazon Lightsail vs EC2 systems) have different networking requirements, so we'd have to maintain custom installers at all providers to support those systems, and that's not something we are looking to do at this point.

For example, less experienced admins have let us know that setting up a Google Cloud server is too complicated. I personally can get one going in a few minutes, but there are some steps that just have to be done over SSH, such as changing the root password, setting up the hostname, configuring the hostname fix script so it doesn't change on reboots, editing the static IP configuration, etc. So, do we work to automate that for every single type of Google instance, or do we hope users that are at the point of needing a dedicated machine but aren't so technical find their own solution? There isn't really a great answer to that to make all the sides happy.

I wish the configuration options were easier as well, but with the cloud/virtualization technologies changing so quickly, there really isn't a great "one click" install platform.

 

[clopezi]

@andrew.n - we're definitely aware of issues people run into when installing CentOS 7 and CentOS 8. The problem for us is that each hosting provider, and in some cases even products under the same host (such as Amazon Lightsail vs EC2 systems) have different networking requirements, so we'd have to maintain custom installers at all providers to support those systems, and that's not something we are looking to do at this point.

For example, less experienced admins have let us know that setting up a Google Cloud server is too complicated. I personally can get one going in a few minutes, but there are some steps that just have to be done over SSH, such as changing the root password, setting up the hostname, configuring the hostname fix script so it doesn't change on reboots, editing the static IP configuration, etc. So, do we work to automate that for every single type of Google instance, or do we hope users that are at the point of needing a dedicated machine but aren't so technical find their own solution? There isn't really a great answer to that to make all the sides happy.

I wish the configuration options were easier as well, but with the cloud/virtualization technologies changing so quickly, there really isn't a great "one click" install platform.

I understand it perfectly, it's almost impossible to maintain everything, but in this case, it's a dedicated server, bare metal, so I belive that it's easier to maintain it. However, I hope that OVH update their cPanel template with some RHEL 8 distro.

On the other hand, I'm happy that you are aware of this problems, because yesterday in ticket support, it seemed that I was the only one and was the first time this happened.

King regards,

 

[andrew.n]

@cPRex I do understand that. We experienced less network related issues with CentOS 7 than with AlmaLinux so far though but since it's pretty new the providers are probably not yet prepared for it either plus AlmaLinux became available basically anywhere by converting CentOS 8 which could lead to further issues. As an example is that converting CentOS 8 with cPanel to AlmaLinux is not yet supported however converting CentOS 8 to AlmaLinux and then installing cPanel works.

 

[Nathan Gerard]

Hello,

We are seeing similar issues while installing cPanel on a server running CentOS 8 in OVH.

We have figured out, so I want to share the solution just in case someone found it useful:

1) Edit /run/cloud-init/instance-data.json an remove the IPv4 DHCP settings
2) Run ip addr to see what interfaces aire in use for IPv4 and IPv6, in our case was enp5s0f0
3) Remove old interfaces without use like eth0
4) Edit /etc/sysconfig/network-scripts/ifcfg-enp5s0f0 and set something like this:

DEFROUTE=yes
DEVICE=enp5s0f0
HWADDR=a0:42:3f:...
IPV6ADDR=2001:41d0...
IPV6INIT=yes
IPV6_AUTOCONF=no
IPV6_DEFAULTGW=2001:41d0:...
IPV6_FORCE_ACCEPT_RA=no
ONBOOT=yes
TYPE=Ethernet
USERCTL=no
BOOTPROTO=static
IPADDR=54.xxx
NETMASK=255.255.255.0
GATEWAY=54.xxx.254 (in OVH, same like IPv4 with 254 at end)
MTU=1500
STARTMODE=auto
IPV4_FAILURE_FATAL=yes

We still have some errors, like "RTNETLINK answers: File exists" on systemctl status network, but it's working.

However, cPanel should support OVH configuration out of the box, I'm sure that a important number of cPanel users are on OVH...

Hi,

We tried installing cPanel on a server running CentOS 8 in OVH and faced similar issues.


# service network.service status
Redirecting to /bin/systemctl status network.service
● network.service - LSB: Bring up/down networking
Loaded: loaded (/etc/rc.d/init.d/network; generated)
Active: failed (Result: exit-code) since Thu 2021-07-29 05:53:34 UTC; 12min ago
Docs: man:systemd-sysv-generator(8)
Process: 10142 ExecStart=/etc/rc.d/init.d/network start (code=exited, status=1/FAILURE)
Tasks: 1 (limit: 822233)
Memory: 8.9M
CGroup: /system.slice/network.service
└─2390 /sbin/dhclient -1 -q -lf /var/lib/dhclient/dhclient--enp97s0f0.lease -pf /run/dhclient-enp97s0f0.pid -H lms enp97s0f0

Jul 29 05:53:34 lms.ifp-learning.com network[10142]: RTNETLINK answers: File exists
Jul 29 05:53:34 lms.ifp-learning.com network[10142]: RTNETLINK answers: File exists
Jul 29 05:53:34 lms.ifp-learning.com network[10142]: RTNETLINK answers: File exists
Jul 29 05:53:34 lms.ifp-learning.com network[10142]: RTNETLINK answers: File exists
Jul 29 05:53:34 lms.ifp-learning.com network[10142]: RTNETLINK answers: File exists
Jul 29 05:53:34 lms.ifp-learning.com network[10142]: RTNETLINK answers: File exists
Jul 29 05:53:34 lms.ifp-learning.com network[10142]: RTNETLINK answers: File exists
Jul 29 05:53:34 lms.ifp-learning.com systemd[1]: network.service: Control process exited, code=exited status=1
Jul 29 05:53:34 lms.ifp-learning.com systemd[1]: network.service: Failed with result 'exit-code'.
Jul 29 05:53:34 lms.ifp-learning.com systemd[1]: Failed to start LSB: Bring up/down networking.


Were you able to find a permenant fix for this or did OVH offer any support? Else can you please explain the lines you have commented out in run/cloud-init/instance-data.json. Also will these changes persist after a reboot?

 

[eadile]

Hello @Nathan Gerard,
We are confronting the same problem with our Bare metal server and Cpanel, did you found a solution ?
Thank you.

 

[andrew.n]

@Nathan Gerard and @eadile have you tried what @clopezi mentioned?

 

[eadile]

@Nathan Gerard and @eadile have you tried what @clopezi mentioned?

Tried, not working

 

[andrew.n]

What is the error you are getting or simply the machine loses network after a reboot? are you able to manually bring up eth0 after a reboot?