Host unable to reach container intermittently

[KarlB]

I'm running Proxmox 4.3-1/e7cdc165.

I have a host, 'Marvin', running Proxmox and several containers.

Since this is in our office, as a development environment, we have a single external IP address. I use Apache on Marvin to proxy specific sites to specific containers.

The issue is that Marvin seems to keep losing communication with one of the containers and I can't see why. I have disabled firewalls (UFW) on both, to no effect. I have restarted services, to no effect.

The problem simply comes and goes with no obvious cause either way.

In the below outputs, I have replaced the specific site name to protect the identify of my organisation. Also, I'm not actually allowed to post any links, so I've had to butcher the output somewhat. I have instead replaced the site url with #URL#.

When the problem is in effect, here is the output from curl:

karl@marvin:~$ curl -v #URL#
* Rebuilt URL to: #URL#
* Hostname was NOT found in DNS cache
*   Trying 192.168.1.42...
* connect to 192.168.1.42 port 80 failed: Connection refused
* Failed to connect to #URL# port 80: Connection refused
* Closing connection 0
curl: (7) Failed to connect to #URL# port 80: Connection refused

Port 443 is similarly affected.

Interestingly, connecting to SSH works and pings work:
karl@marvin:~$ sudo ping #URL#
PING #URL# (192.168.1.42) 56(84) bytes of data.
64 bytes from #URL# (192.168.1.42): icmp_seq=1 ttl=63 time=0.721 ms

Other containers, and my own workstation can connect to Apache on the container on ports 80 and 443. There is nothing shown in the apache logs on the container to indicate that requests are being made from Marvin.

Does anyone have any idea what could be causing this, or how to investigate further?

 

[Richard]

When the problem is in effect, here is the output from curl:

karl@marvin:~$ curl -v #URL#
* Rebuilt URL to: #URL#
* Hostname was NOT found in DNS cache
*   Trying 192.168.1.42...
* connect to 192.168.1.42 port 80 failed: Connection refused
* Failed to connect to #URL# port 80: Connection refused
* Closing connection 0
curl: (7) Failed to connect to #URL# port 80: Connection refused

Port 443 is similarly affected.

Interestingly, connecting to SSH works and pings work:
karl@marvin:~$ sudo ping #URL#
PING #URL# (192.168.1.42) 56(84) bytes of data.
64 bytes from #URL# (192.168.1.42): icmp_seq=1 ttl=63 time=0.721 ms

Other containers, and my own workstation can connect to Apache on the container on ports 80 and 443. There is nothing shown in the apache logs on the container to indicate that requests are being made from Marvin.

Look like the ports are not binded to all (0.0.0.0) addresses - what reports

in the above mentioned VM 192.168.1.42
netstat -a -n | grep "80\|443"

(if it is a LINUX machine)?

 

[KarlB]

I can try that when the error occurs again, but I was able to connect to port 80/443 on 192.168.1.42 from other servers/workstations (on the same interface) without any problems.