MAC learning for cross-site layer-2 connectivity
Mac learning for cross-site layer-2 connectivity. Some embodiments provide a method for enabling data traffic to be sent between across which a logical network spans. A logical network segment (i.e., a layer-2 network segment such as a logical switch) spans at least two sites via tunnels between network elements deployed in the sites that operate as bridges (or 2-port switches) to pass data traffic between the sites without leaving the logical network segment. Because these network elements operate similarly to bridges, they are not assigned network (layer 3) addresses (e.g., IP addresses). In order to send address resolution protocol (ARP) requests to a router in one site, the network element in another site uses a network address of one of the data compute nodes (DCNs) operating in its site that is connected to the logical network segment spanning the two sites.
1. A method for sending data traffic from a first site to a router in a second site, the method comprising:
at a first network element, located at the first site, that provides a direct connection to the second site to enable a logical network to span both the first and second sites, wherein the first network element does not have an assigned network address;
identifying a data compute node (DCN), located at the first site, that connects to a logical network segment spanning the first and second sites;
using a network address of the identified DCN to send an address resolution protocol (ARP) request to the router located at the second site, the network address being borrowed, the ARP request being sent across a tunnel;
detecting migration of the DCN to the second site and selecting, at random, using at least a randomization function, a new DCN at the first site for borrowing a network address to continue ARP transmissions; and
upon receiving an ARP response with a data link address of the router located at the second site, using the data link address to forward data traffic directed to a subset of DCNs at the second site through the direct connection to the router located at the second site;
wherein the data traffic is intercepted by virtualization software of a host on which a source DCN operates and is directed to the first network element instead of being sent to an implementation of a second router executing on the host.
2. The method of claim 1 , wherein the first network element provides the direct connection to the second site via a tunnel to a corresponding second network element located at the second site.
3. The method of claim 2 , wherein the logical network segment spanning the first and second sites enables migration of DCNs between the first and second sites while the DCNs remain operational.
4. The method of claim 1 , wherein the router at the second site is a first router and the second router at the first site shares a network address with the first router.
5. The method of claim 4 , wherein the second router is a distributed logical router that is implemented in the virtualization software of a set of host computers at the first site, including a particular host computer on which the identified DCN operates.
6. The method of claim 5 , wherein:
the data traffic directed to the subset of DCNs at the second site comprises a data message sent from the source DCN at the first site to a network address of the first and second routers.
7. The method of claim 4 , wherein the first and second routers are both connected to the logical network segment that spans the first and second sites.
8. The method of claim 7 , wherein:
the logical network segment is a first logical network segment, wherein the first router also connects to a second logical network segment that is confined to the second site; and
the first network element uses the data link address of the first router to forward data traffic sent from DCNs at the first site that are connected to the first logical network segment and directed to DCNs at the second site that are connected to the second logical network segment.
9. The method of claim 1 , wherein the identified DCN is a first DCN, the method further comprising:
detecting that the first DCN has been migrated to the second site;
identifying a second DCN at the first site that connects to the first logical network segment; and
using a network address of the second DCN rather than the network address of the first DCN to send an ARP request to the router located at the second site.
10. The method of claim 9 , wherein after migration the first DCN communicates directly with the router located at the second site as a default gateway.
11. The method of claim 1 , wherein the first site is a virtual datacenter implemented in a cloud that extends the second site and the second site is a physical on-premises datacenter.
12. The method of claim 11 , wherein the first and second sites are managed by a same enterprise.
13. A non-transitory machine-readable medium storing a first network element, located at a first site, which when executed by at least one processing unit sends data traffic from a first site to a router in a second site, wherein the first network element (i) provides a direct connection to the second site to enable a logical network to span both the first and second sites and (ii) does not have an assigned network address, the first network element comprising sets of instructions for:
identifying a data compute node (DCN), located at the first site, that connects to a logical network segment spanning the first and second sites;
using a network address of the identified DCN to send an address resolution protocol (ARP) request to the router located at the second site, the network address being borrowed, the ARP request being sent across a tunnel;
detecting migration of the DCN to the second site and selecting, at random, using at least a randomization function, a new DCN at the first site for borrowing a network address to continue ARP transmissions; and
upon receiving an ARP response with a data link address of the router located at the second site, using the data link address to forward data traffic directed to a subset of DCNs at the second site through the direct connection to the router located at the second site;
wherein the data traffic is intercepted by virtualization software of a host on which a source DCN operates and is directed to the first network element instead of being sent to an implementation of a second router executing on the host.
14. The non-transitory machine-readable medium of claim 13 , wherein the first network element provides the direct connection to the second site via a tunnel to a corresponding second network element located at the second site.
15. The non-transitory machine-readable medium of claim 14 , wherein the logical network segment spanning the first and second sites enables migration of DCNs between the first and second sites while the DCNs remain operational.
16. The non-transitory machine-readable medium of claim 13 , wherein the router at the second site is a first router and a second router at the first site shares a network address with the first router.
17. The non-transitory machine-readable medium of claim 16 , wherein:
the second router is a distributed logical router that is implemented in virtualization software of a set of host computers at the first site, including a particular host computer on which the identified DCN operates;
the data traffic directed to the subset of DCNs at the second site comprises a data message sent from the source DCN at the first site to a network address of the first and second routers.
18. The non-transitory machine-readable medium of claim 16 , wherein:
the logical network segment is a first logical network segment;
the first and second routers are both connected to the first logical network segment;
the first router also connects to a second logical network segment that is confined to the second site; and
the first network element uses the data link address of the first router to forward data traffic sent from DCNs at the first site that are connected to the first logical network segment and directed to DCNs at the second site that are connected to the second logical network segment.
19. The non-transitory machine-readable medium of claim 13 , wherein the identified DCN is a first DCN, the program further comprising sets of instructions for:
detecting that the first DCN has been migrated to the second site;
identifying a second DCN at the first site that connects to the first logical network segment; and
using a network address of the second DCN rather than the network address of the first DCN to send an ARP request to the router located at the second site.
20. The non-transitory machine-readable medium of claim 19 , wherein after migration the first DCN communicates directly with the router located at the second site as a default gateway.
21. The non-transitory machine-readable medium of claim 13 , wherein the first site is a virtual datacenter implemented in a cloud that extends the second site and the second site is a physical on-premises datacenter.