Sensational title but an apt one for this blog post as it seems to be frequently asked question on the #cloudstack IRC channel and on the mailing lists. 🙂
Below are a few troubleshooting steps to check whats going wrong with your network…
- The switches have to be configured correctly to pass VLAN traffic. You can verify if VLAN traffic is working by bringing up a tagged interface on the hosts and pinging between them as below…
# Host 1 $ vconfig add eth0 64 $ ifconfig eth0.64 1.2.3.4 netmask 255.255.255.0 up $ ping 1.2.3.5 # host 2 $ vconfig add eth0 64 $ ifconfig eth0.64 1.2.3.5 netmask 255.255.255.0 up $ ping 1.2.3.4
If the pings dont work, run tcpdump(8) all over the place to check who is gobbling up the packets. Ultimately, if the switches are not configured correctly, CloudStack networking wont work so fix the physical networking issues before you proceed to the next steps.
- Ensure Traffic Labels are set for the Zone.
Traffic labels need to be set for all hypervisors including XenServer, KVM and VMware types. You can configure traffic labels when you creating a new zone using the Add Zone Wizard.
On an existing zone, you can modify the traffic labels by going to Infrastructure, Zones, Physical Network tab.
[root@csman1-1 ~]# cloudmonkey list traffictypes physicalnetworkid=41cb7ff6-8eb2-4630-b577-1da25e0e1145 count = 4 traffictype: id = cd0915fe-a660-4a82-9df7-34aebf90003e kvmnetworklabel = cloudbr0 physicalnetworkid = 41cb7ff6-8eb2-4630-b577-1da25e0e1145 traffictype = Guest xennetworklabel = MGMT ======================================================== id = f5524b8f-6605-41e4-a982-81a356b2a196 kvmnetworklabel = cloudbr0 physicalnetworkid = 41cb7ff6-8eb2-4630-b577-1da25e0e1145 traffictype = Management xennetworklabel = MGMT ======================================================== id = 266bad0e-7b68-4242-b3ad-f59739346cfd kvmnetworklabel = cloudbr0 physicalnetworkid = 41cb7ff6-8eb2-4630-b577-1da25e0e1145 traffictype = Public xennetworklabel = MGMT ======================================================== id = a2baad4f-7ce7-45a8-9caf-a0b9240adf04 kvmnetworklabel = cloudbr0 physicalnetworkid = 41cb7ff6-8eb2-4630-b577-1da25e0e1145 traffictype = Storage xennetworklabel = MGMT =========================================================
- KVM traffic labels require to be named as “cloudbr0”, “cloudbr2”, “cloudbrN” etc and the corresponding bridge must exist on the KVM hosts. If you create labels/bridges with any other names, CloudStack (atleast earlier versions did) seems to ignore them. CloudStack does not create the physical bridges on the KVM hosts, you need to create them before before adding the host to Cloudstack.
[root@kvm1-2 ~]# ifconfig cloudbr0 cloudbr0 Link encap:Ethernet HWaddr 00:0C:29:EF:7D:78 inet addr:192.168.44.22 Bcast:192.168.44.255 Mask:255.255.255.0 inet6 addr: fe80::20c:29ff:feef:7d78/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:92435 errors:0 dropped:0 overruns:0 frame:0 TX packets:50596 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes:94985932 (90.5 MiB) TX bytes:61635793 (58.7 MiB) - The Virtual Router, SSVM, CPVM public interface would be bridged to a physical interface on the host. In the example below, cloudbr0 is the public interface and CloudStack has correctly created the virtual interfaces bridge. This virtual interface to physical interface mapping is done automatically by CloudStack using the traffic label settings for the Zone. If you have provided correct settings and still dont have a working working Internet, check the switching layer before you debug any further. You can verify traffic using tcpdump on the virtual, physical and bridge interfaces.
[root@kvm1-2 ~]# brctl show bridge name bridge id STP enabled interfaces breth0-64 8000.000c29ef7d78 no eth0.64 vnet2 cloud0 8000.fe00a9fe0219 no vnet0 cloudbr0 8000.000c29ef7d78 no eth0 vnet1 vnet3 virbr0 8000.5254008e321a yes virbr0-nic
[root@vxen1-1 ~]# brctl show bridge name bridge id STP enabled interfaces xapi0 0000.e2b76d0a1149 no vif1.0 xenbr0 0000.000c299b54dc no eth0 xapi1 vif1.1 vif1.2
- Pre-create labels on the XenServer Hosts. Similar to KVM bridge setup, traffic labels must also be pre-created on the XenServer hosts before adding them to CloudStack.
[root@vxen1-1 ~]# xe network-list uuid ( RO) : aaa-bbb-ccc-ddd name-label ( RW): MGMT name-description ( RW): bridge ( RO): xenbr0 - The Internet would be accessible from both the SSVM and CPVM instances by default. Their public IPs will also be directly pingable from the Internet. Please note that these test would work only if your switches and traffic labels are configured correctly for your environment. If your SSVM/CPVM cant reach the Internet, its very unlikely that the Virtual Router (VR) can also the reach the Internet suggesting that its either a switching issue or incorrectly assigned traffic labels. Fix the SSVM/CPVM issues before you debug VR issues.
root@s-1-VM:~# ping -c 3 google.com PING google.com (74.125.236.164): 56 data bytes 64 bytes from 74.125.236.164: icmp_seq=0 ttl=55 time=26.932 ms 64 bytes from 74.125.236.164: icmp_seq=1 ttl=55 time=29.156 ms 64 bytes from 74.125.236.164: icmp_seq=2 ttl=55 time=25.000 ms --- google.com ping statistics --- 3 packets transmitted, 3 packets received, 0% packet loss round-trip min/avg/max/stddev = 25.000/27.029/29.156/1.698 ms
root@v-2-VM:~# ping -c 3 google.com PING google.com (74.125.236.164): 56 data bytes 64 bytes from 74.125.236.164: icmp_seq=0 ttl=55 time=32.125 ms 64 bytes from 74.125.236.164: icmp_seq=1 ttl=55 time=26.324 ms 64 bytes from 74.125.236.164: icmp_seq=2 ttl=55 time=37.001 ms --- google.com ping statistics --- 3 packets transmitted, 3 packets received, 0% packet loss round-trip min/avg/max/stddev = 26.324/31.817/37.001/4.364 ms
- The Virtual Router (VR) should also be able to reach the Internet without having any Egress rules. The Egress rules only control forwarded traffic and not traffic that originates on the VR itself.
root@r-4-VM:~# ping -c 3 google.com PING google.com (74.125.236.164): 56 data bytes 64 bytes from 74.125.236.164: icmp_seq=0 ttl=55 time=28.098 ms 64 bytes from 74.125.236.164: icmp_seq=1 ttl=55 time=34.785 ms 64 bytes from 74.125.236.164: icmp_seq=2 ttl=55 time=69.179 ms --- google.com ping statistics --- 3 packets transmitted, 3 packets received, 0% packet loss round-trip min/avg/max/stddev = 28.098/44.021/69.179/17.998 ms
- However, the Virtual Router’s (VR) Source NAT Public IP address WONT be reachable until appropriate Ingress rules are in place. You can add Ingress rules under Network, Guest Network, IP Address, Firewall setting page.
- The VM Instances by default wont be able to access the Internet. Add Egress rules to permit traffic.
- Some users have reported that flushing IPTables rules (or changing routes) on the SSVM, CPVM or the Virtual Router makes the Internet work. This is not expected behaviour and suggests that your networking settings are incorrect. No IPtables/route changes are required on the SSVM, CPVM or the VR. Go back and double check all your settings.
In a vast majority of the cases, the problem has turned out to be at the switching layer where the L3 switches were configured incorrectly.
Shanker Balan
Shanker Balan is a devops and infrastructure freelancer with over 14 years of industry experience in large scale Internet systems. He is available for both short term and long term projects on contract. Please use the Contact Form for any enquiry.
Follow Me:
Hi, Shanker,
Could your provide an example about how to create physical bridges on a Ubuntu KVM node? How many bridges should be created?
Hi Dan,
I dont have a KVM setup anymore or a Ubuntu system nearby. Sorry for not being of much help.
Hi Shanker. Your articles have been very helpful during my Cloudstack implementation. Example #4 above is happening to me, and I’m not quite sure how to resolve it. I’ve removed the host from Cloudstack and re-added it with no results. The bridged interface (on my host) is named brp3p1-276. I have a VPC and a guest network created (called Web Tier). It doesn’t look like there’s a file in CentOS called brp3p1-276 on my working host. I’m out of ideas. Thoughts?