Hi, Vadim, Thanks for the hints. Yes, the route table looks weird, but this comes from the official guide of network set up of cloudstack: http://docs.cloudstack.apache.org/projects/cloudstack-installation/en/latest/hypervisor/kvm.html
So I confused if the network pattern from the official guide is misleading or one should create VLANs one by one manually and set different IP address for each of them explicitly(eth0.100,eth0.200 and eth0.300)? Note that my network is nothing special, just put the Management Server, the KVM hypervisor and guest VMs all in the 192.168.0.* network. Could you suggest an simple example of how the file "/etc/network/interfaces" on KVM host looks like? The official pattern is as below but certainly does not work: =================== auto lo iface lo inet loopback # The primary network interface auto eth0.100 iface eth0.100 inet static address 192.168.42.11 netmask 255.255.255.240 gateway 192.168.42.1 dns-nameservers 8.8.8.8 8.8.4.4 dns-domain lab.example.org # Public network auto cloudbr0 iface cloudbr0 inet manual bridge_ports eth0.200 bridge_fd 5 bridge_stp off bridge_maxwait 1 # Private network auto cloudbr1 iface cloudbr1 inet manual bridge_ports eth0.300 bridge_fd 5 bridge_stp off bridge_maxwait 1 ==================== Cheers, Dan 2014-11-23 7:51 GMT-06:00 Vadim Kimlaychuk <vadim.kimlayc...@elion.ee>: > Hello Dan, > > Sorry for the late reply, have to be out from internet sometimes > :) > You have problem with routing. According to arp table default > interface -- eth2 it is not connected to you 192.168.0.* network and can't > see gateway (192.168.0.100) what is not ture. Arp table looks ugly -- you > have 3 interfaces connected to the same network with same Metric. How do > you think VM will choose the correct one?? I suppose it will take the > first one in line -- eth1 and that is why you see eth1 with complete MAC > for 192.168.0.100. It always uses eth1 to reach other networks, but default > gw is eth2. > For me, network layout looks awkward. It is better to > re-partition and re-think it. If you don't want to -- try to fix routing > table manually. Put metrics to your routes and eth2 should be the highest > (i.e. have lowest number). This will temporarty solve the problem, but you > still have a chance to loose your changes after reboot, because routing > table is created dynamically. > > Fix you network and I believe everything will work out-of-the box. > > Regards, > > Vadim. > > ________________________________________ > From: Dan Dong [dongda...@gmail.com] > Sent: Wednesday, November 19, 2014 17:45 > To: users@cloudstack.apache.org > Subject: Re: To let SSVM reach outside network. > > Hi, Vadim, > 1. route -n > root@s-1-VM:~# route -n > Kernel IP routing table > Destination Gateway Genmask Flags Metric Ref Use > Iface > 0.0.0.0 192.168.0.100 0.0.0.0 UG 0 0 0 > eth2 > 169.254.0.0 0.0.0.0 255.255.0.0 U 0 0 0 > eth0 > 192.168.0.0 0.0.0.0 255.255.255.0 U 0 0 0 > eth1 > 192.168.0.0 0.0.0.0 255.255.255.0 U 0 0 0 > eth2 > 192.168.0.0 0.0.0.0 255.255.255.0 U 0 0 0 > eth3 > > Noted that to the same 192.168.0.0 network, eth1 is searched before eth2, > while eth2 is supposed to be the public NIC. Should one change the order? > > 2. root@s-1-VM:~# arp -n > cloud (192.168.0.100) at 84:2b:2b:01:c3:d0 [ether] on eth1 > cloud (192.168.0.100) at <incomplete> on eth2 > ? (169.254.0.1) at fe:00:a9:fe:01:7a [ether] on eth0 > > Noted that MAC is incomplete on eth2, while MAC detected on eth1 is the > correct address of the internal NIC of the Management Server(the gateway > for 192.168.0.0 subnet). > > 3. The network layout is quite simple here( basic network): > Management Server external NIC:10.* > Management Server internal NIC: 192.168.0.100 > > KVM host NIC: 192.168.0.101 > SSVM and guest VMs are all in 192.168.0.*/24 network. > > 4. NAT is configured on the Management Server so 192.168.0.0/24 subnet > could reach the internet(download packages from internet etc., e.g, from > KVM host itself) > > Cheers, > Dan > > > 2014-11-19 1:23 GMT-06:00 Vadim Kimlaychuk <vadim.kimlayc...@elion.ee>: > > > Dan, > > > > I would suggest you to use external proxy/GW to hide you network. Not the > > same host that contains management server. Because if you would like to > > scale it up - how do you do it?? Management server should be rather > simple > > component of infrastructure and to put more functions on it is a bad > idea. > > > > Anyway > > 1. what does "route -n" says on your SSVM ?? > > 2. If you use KVM your bridges tag the traffic. Are you sure that it is > > properly handled on your switch?? Run "arp -a" at your SSVM and see if VM > > "knowns" MAC of the gateway. > > > > I do expect you put the output of route and arp here otherwise we will > > continue guessing. > > > > It would be even better if you describe your network layout like on the > > first picture here: > > > http://cloudstack-administration.readthedocs.org/en/latest/networking_and_traffic.html > > > > Vadim. > > > > -----Original Message----- > > From: Dan Dong [mailto:dongda...@gmail.com] > > Sent: Tuesday, November 18, 2014 5:23 PM > > To: users@cloudstack.apache.org > > Subject: Re: To let SSVM reach outside network. > > > > Hi, Vadim, > > We have to use 2 NICs on the management server as we want to hide the > > cloudstack cluster behind the 10.* network, so all KVM hosts and guest > VMs > > are in the 192.168.0.* subnet, and they connect to the management > server's > > internal NIC(192.168.0.100). Is it a rule that the management server can > > only use one NIC? And the KVM hypersior host can reach the internet and > > download packages, but SSVM running on it could not see the internet. > > > > Cheers, > > Dan > > > > > > 2014-11-18 3:20 GMT-06:00 Vadim Kimlaychuk <vadim.kimlayc...@elion.ee>: > > > > > Hello Dan, > > > > > > It seems there is something wrong with your network setup and > > > here are some places to search: > > > 1. Why your management server has 2 NICs? It should not work as NAT, > > > proxy or any kind of switch - keep this in mind. > > > 2. SSVM normally has to have 3 interfaces (at least). One -- with > > > public IP, one -- management network IP and one -- link local IP. If > > > you have separate storage network -- it may have one more, but this is > > > not your case. Check routing table for your SSVM with "route -n" > > > command. Find your default gateway. It should be public interface. > > > 3. Did you set up KVM hypervisor network correctly? Does it have 2 > > > interfaces like in the setup guide? Does your hypervisor have access > > > to internet to be able to download template? > > > > > > Vadim. > > > > > > -----Original Message----- > > > From: Dan Dong [mailto:dongda...@gmail.com] > > > Sent: Monday, November 17, 2014 7:02 PM > > > To: users@cloudstack.apache.org > > > Subject: To let SSVM reach outside network. > > > > > > Hi, All, > > > I found I could not register my ISO image to the cloudstack( through > > > URL of > > http://releases.ubuntu.com/14.04.1/ubuntu-14.04.1-server-amd64.iso). > > > The SSVM is running and health check show no problems on it. But it > > > just could not see the outside network, although SSVM is on the same > > > network as KVM host, i.e. 192.168.0.*. My setup is as following: > > > > > > Management Server external NIC:10.* > > > Management Server internal NIC: 192.168.0.100 > > > > > > KVM host NIC: 192.168.0.101 > > > guest VMs are all in 192.168.0.*/24 network. > > > > > > From SSVM I can ping the internal IP of the Management Server at > > > 192.168.0.100, but could not ping the external IP of it at 10.*. > > > > > > From KVM host itself I can reach the outside internet as NAT is > > > configured on the Management Server to let the 192.168.0.0 traffic > > > through. Any hints how to let SSVM to reach the internet? > > > > > > > > > Cheers, > > > Dan > > > > > >