Web UI times out after install

[isaacngym]

Somehow my proxmox install borked itself, so i tried reinstalling 2 days ago.

My router is configured not to assign IPs outside 192.168.72.4 to 192.168.72.180. Used Balena to image 8.2-1 to a thumbdrive, booted into it, installed it. all good. Then I configured the installer to use 192.168.192/27, so that's 192-223. Installation finished, rebooted, was instructed to go to 192.168.72.192:8006.

https://192.168.72.192:8006 timed out on me.

Thinking i did something wrong, I reinstalled proxmox again. Still the same result. I tried to ssh into my proxmox box from another laptop, only to find that that didn't work (SSH used to work fine).

SSH from proxmox to my other linux machine works fine. Both machines can ping one another.

running `wget --no-check-certificate 192.168.72.192:8006" on the proxmox machine works. Running it from my other linux box does not.

nmap shows that port 8006 is not open.

running iptables -L shows this:

root@pve:~# iptables -L
Chain INPUT (policy ACCEPT)
target prot opt source destination


Chain FORWARD (policy ACCEPT)
target prot opt source destination


Chain OUTPUT (policy ACCEPT)
target prot opt source destination

and pveproxy status says running.

I'm puzzled - when I installed tailscale on my device on one of my reinstalls, and exposed localhost:8006 via tailscale serve i could access the admin console. good ol’ ssh worked fine when i would try connect to the tailscale IP address.


no idea what else the issue is - could anyone help me?

 

[esi_y]

Somehow my proxmox install borked itself, so i tried reinstalling 2 days ago.

This was because of the tailscale? (Reading from bottom up.)

My router is configured not to assign IPs outside 192.168.72.4 to 192.168.72.180. Used Balena to image 8.2-1 to a thumbdrive, booted into it, installed it. all good. Then I configured the installer to use 192.168.192/27, so that's 192-223. Installation finished, rebooted, was instructed to go to 192.168.72.192:8006.

Your 192.168.72.192/27 - I believe that's .193 to .222 for hosts. .192 is network address.

If you like these numberings, I suggest doing 10.x.y.z networks instead.

PS You can actually use DHCP too: https://forum.proxmox.com/threads/d...-only-no-static-reservation-necessary.155125/

 

[bbgeek17]

My router is configured not to assign IPs outside 192.168.72.4 to 192.168.72.180.

Is this because you are assigning IPs statically from that range?

192.168.192/27

Is your router configured to understand routing between 192.168.72 and 192.168.192 ?

It seems to me that you overcomplicated your network. Also some essential information is missing, i.e. netmask of 192.168.72 and how the two networks are supposed to talk to each other.

As usual, start with basics - check via curl against "localhost" and local IP whether PVE services are up directly from PVE.
Move on to checks from the host in the same network. Continue with more complex routing checks.

In the end, none of this is PVE specific or related to "Proxmox VE: Installation and configuration" , its basic networking and Linux management.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[esi_y]

Is your router configured to understand routing between 192.168.72 and 192.168.192 ?

I believe that was just a typo from the OP, it's all 192.168.72.192/27.

 

[bbgeek17]

I believe that was just a typo from the OP, it's all 192.168.72.192/27.

Anything is possible. In that case as you pointed out x.y.z.192 is the network address and not a valid setting for PVE's IP.
Additionally, we dont know what OP's client IP is, if it is outside the specified /27 range, is the router able to handle it.

In any case, if this is a home setup - its overcomplicated testing ground. Op should drop back to using /24 and make his network experiments in VMs.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[esi_y]

Additionally, we dont know what OP's client IP is, if it is outside the specified /27 range, is the router able to handle it.

I believe, as you ...

In any case, if this is a home setup

... he is doing all the hocus pocus just to satisfy having DHCP and some tiny safe-from-DHCP portion on the same route. Most home routers would not do custom routing tables.

Op should drop back to using /24 and make his network experiments in VMs.

But this is a genuine problem of the Installer, it e.g. allows you to set 10.10.10.0/24 just fine, should not be happening.

EDIT: Filed as https://bugzilla.proxmox.com/show_bug.cgi?id=5757

 

[isaacngym]

... he is doing all the hocus pocus just to satisfy having DHCP and some tiny safe-from-DHCP portion on the same route. Most home routers would not do custom routing tables.

yes, that's exactly what happened. I read that proxmox really doesn't like DHCP so i saved some IP addresses from being assigned by the router and thought proxmox would use them. Previously when I did my install i didn't know as much about routing but i suspect my network looking different back then saved me. (my room had a router separate from the rest of my family, in a different subnet)

I reinstalled proxmox using 192.168.72.248/24 and can reach the Web UI now.

it e.g. allows you to set 10.10.10.0/24 just fine, should not be happening.

Let me test my understanding here - the installer is not asking what IP range i want proxmox and its CT/VMs to be created within; it's asking me what IP address i want the proxmox host to be at; and which IPs are in the rest of the subnet.

When I tried to access 192.168.72.192:8006 or attempted to SSH in, the proxmox host received the packets, looked at the /27 netmask and said "huh, weird" and didn't bother replying because it came from outside its subnet?

Your example of 10.10.10.0/24 is invalid because proxmox can't use 10.10.10.0 as its own IP; it's not asking for the CIDR range for the "VPC" that proxmox is in charge of, it's asking for what IP does proxmox want and what sized subnet does it belong to?


Appreciate the help! There's still the problem of restoring my VMs from their backups, but i'll make another thread for that.

 

[esi_y]

yes, that's exactly what happened. I read that proxmox really doesn't like DHCP

I do not really like to use terms like "like" with equipment that is meant to compute everything.

PVE actually does not care if you assign it address via DHCP, just its installer pretends it is the only way (did you ever notice the installer itself uses DHCP, only after reboot it's down to static config?) - see also the linked threads above.

PVE simply uses its hostname-to-IP-address mapping as a mechanism to find out its own IP address (strange way of doing it), i.e. it knows its hostname and wants to be able to determine what IP it goes by in the network from the point of view of the others. DNS solves that problem quite well, but someone at Proxmox prefers wrong way of doing it. Also, the problem does not really exist for single node, as it does not really need to care what address it has, but codebase sharing and sloppiness ...

You can get a bit of glimpse on this in the post (just that one piece on 127.0.1.1) here:
https://forum.proxmox.com/threads/dhcp-cluster-deployment.154780/#post-705210

so i saved some IP addresses from being assigned by the router and thought proxmox would use them.

I think you figured by now that DHCP pool can only serve some of the addresses of the subnet, you do not have to have DHCP only and static subnets. Some people even set fake DHCP reservations despite they are - obviosly - never given out to anyone if the machine is statically configured.

But I have no idea what other things you might have intended with that /27.

Let me test my understanding here - the installer is not asking what IP range i want proxmox and its CT/VMs to be created within;

Correct, it is NOT. But see more below.

it's asking me what IP address i want the proxmox host to be at; and which IPs are in the rest of the subnet.

Exactly. The latter is an unusual way to put it, but yes, you need to specify the address in the network (subnet) of which the host (PVE) is part of. What you will do with the guests is entirely up to you later. The standard install leaves you with a bridge on that interface (of which you can think as a software switch).

When I tried to access 192.168.72.192:8006 or attempted to SSH in, the proxmox host received the packets, looked at the /27 netmask and said "huh, weird" and didn't bother replying because it came from outside its subnet?

I think something else happened, but it would really be interesting to see the topology you have / had there to answer it properly. Usually if you have one home router that allows you to set up multiple private subnets, it would route between them just fine (if e.g. your machine accessing was on one and PVE on another) - they are "connected" routes after all.

The bizzare part is that you were sending traffic to the so-called network address. You can't really use 2 addresses in a network (subnet), the first and the last, first being the network address and last reserved for broadcast. (Ok, I lie here a bit, there's even /31 mask possible.) Note this has nothing to do with nice numbers, you actually can have a host address with 0 at the end (lots of people omit it out of habit), if you e.g. have 10.10.0.0/16 network, then then 10.10.0.0 is the network address, 10.10.255.255 is the broadcast for it, but e.g. 10.10.10.0 is a valid address that could be used for a host.

Now I am just guessing whatever was actually working in your case was working on the link layer (one below network layer which routes) as the "router" (also moonlighting as a switch in this case) learned the MAC addresses and simply sent whatever it got to whoever claimed was that address.

Your example of 10.10.10.0/24 is invalid because proxmox can't use 10.10.10.0 as its own IP;

Correct, not on /24, see above.

it's not asking for the CIDR range for the "VPC" that proxmox is in charge of,

I do not know what you are setting up your networking on PVE like, in fact I do not know much about its GUI myself. I know however that fresh install leaves you with one bridge. And since you can think of bridge as of a switch, PVE in that scenario really is not "in charge" of anything, it simply lives with that interface on that subnet and allows its VMs to "plug into" that bridge as if the network card was shared amongst multiple machines.

So by extension, it might occur to you that PVE is somehow in charge of anything, but unless you set it up to e.g. route, etc., it does not do anything like that.

Not sure where you got your VPC term from, but if you liken this to public cloud, then there's no VPC in PVE after stock install, you would need to create it afterwards.

it's asking for what IP does proxmox want and what sized subnet does it belong to?

Yes that's the only thing installer asks for. But it also happens to create that bridge. Basically, the installer makes you be the human DHCP serving staff, one-off, while at that.