epgoraj [Wed, 1 Mar 2017 06:54:03 +0000 (12:24 +0530)]
BUG:7896 OptimisticLockFailedException
Change-Id: I977feeb0226613b1a93cfb8092d4afa88926979e
Signed-off-by: epgoraj <p.govinda.rajulu@ericsson.com>
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>
Vinh Nguyen [Tue, 28 Feb 2017 22:08:03 +0000 (14:08 -0800)]
Bug 7863 - Add Layer 4 Match for flow entries for TCP/UDP security group rule with no min/max
The methods Ingress/EgressAclService::ingress/egressAclTcp/Udp
incorrectly assume that portRangMin/Max values are always specifed.
This is not true when using openstack command lines to
define SG rule for TCP/UDP with no min/max ports specified.
As a result, no layer 4 match are added to the flow entries.
The fix is to add logic to handle null min/max values for the
above scenarios.
Change-Id: I537f51fdd56d8aca3711a38cd5fd2c5eba5d9af9
Signed-off-by: Vinh Nguyen <vinh.nguyen@hcl.com>
eupakir [Sat, 25 Feb 2017 14:08:38 +0000 (19:38 +0530)]
BUG 7727 : Local and Connected routes do not get imported
Route Origin for Subnet Routes and Local VM routes were changed from
STATIC to CONNECTED and LOCAL, respectively. Hence, the import/export RT
functionality for VPNs was broken.
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>
Conflicts:
vpnservice/fibmanager/fibmanager-api/src/main/java/org/opendaylight/netvirt/fibmanager/api/FibHelper.java
vpnservice/fibmanager/fibmanager-impl/src/main/java/org/opendaylight/netvirt/fibmanager/FibUtil.java
vpnservice/fibmanager/fibmanager-impl/src/main/java/org/opendaylight/netvirt/fibmanager/VrfEntryListener.java
vpnservice/vpnmanager/vpnmanager-impl/src/main/java/org/opendaylight/netvirt/vpnmanager/VpnInterfaceManager.java
Change-Id: I04d8326fcb0bddf77934a63c1eb459f23564323b
Signed-off-by: eupakir <kiran.n.upadhyaya@ericsson.com>
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>
Alon Kochba [Wed, 15 Feb 2017 19:16:03 +0000 (21:16 +0200)]
Fix potential NPEs in ELAN tunnel handling
We encountered null ElanInfos in CSIT after deletion
of all the interfaces via REST in BFD Monitoring suite,
this should not prevent other elanDpns from processing.
Conflicts:
vpnservice/elanmanager/elanmanager-impl/src/main/java/org/opendaylight/netvirt/elan/internal/ElanInterfaceManager.java
Change-Id: I465b62cb5133e6aea5326eb4219fc82a70eab70e
Signed-off-by: Alon Kochba <alonko@hpe.com>
Konsta Pozdeev [Thu, 16 Feb 2017 11:46:11 +0000 (13:46 +0200)]
Bug 7418 Run local group creation as async task with key equal to subsequent tasks.
Change-Id: I515b3b874b50f8e53a931e1a3ee3a2e30f877f8e
Signed-off-by: Konsta Pozdeev <konsta.pozdeev@hpe.com>
Somashekar Byrappa [Thu, 2 Feb 2017 14:21:30 +0000 (19:51 +0530)]
Bug 7725: AAP with prefix 0.0.0.0/0 not supported in ACL
+ Added a check. In case of 0.0.0.0/0 prefix, program flow with just "ip"
match skipping the ip-prefix (0.0.0.0/0).
+ This method is called to configure flows both in table 40 and 251.
+ Handled below issue:
- Create VM1 & VM2 with AAP 0.0.0.0/0 and default SG.
- Create VM3 with default SG without AAP.
- Delete VM2. Observe that +new+trk flow configured for VM2's 0.0.0.0/0 AAP
prefix gets deleted (related to remote SG) in 252 table even though VM1 still
exists.
Flow: cookie=0x6900000, duration=4737.343s, table=252, n_packets=0,
n_bytes=0,
priority=61010,ct_state=+new+trk,ip,metadata=0x10000000000/0x1fffff0000000000
actions=ct(commit,zone=5000),resubmit(,220)
+ Earlier flow priority was generated/used based on SG rule UUID. The
above case was not handled with this approach. So, with this fix now
every ct_state=+new+trk flow will have a unique flow priority within table.
Flow priorities are generated based on flow ID.
+ For every switch, two ID pools (one each for ingress and egress) are
maintained which increases the availability of flow priorities.
Change-Id: I646312e1324267913cbb2865a1a8164b23b40372
Signed-off-by: Somashekar Byrappa <somashekar.b@altencalsoftlabs.com>
Signed-off-by: Sam Hague <shague@redhat.com>
Signed-off-by: Somashekar Byrappa <somashekar.b@altencalsoftlabs.com>
eceghkl [Mon, 27 Feb 2017 11:07:37 +0000 (16:37 +0530)]
Fix for GwMac flow deletion during interface delete
* Without the fix, the mac address used for deleting the flow
was obtained after quering interface state again. This
returned null resulting in gwmac table not getting deleted.
* With the fix, interface state is not queried again. Interface
state passed during interface state deletion notification is
used to obtain mac address.
Change-Id: I1c091dc8edb722ab58fdf05f6cba1729a583209e
Signed-off-by: eceghkl <manu.b@ericsson.com>
Abhinav Gupta [Thu, 2 Mar 2017 10:01:37 +0000 (15:31 +0530)]
Bug 7875: Separated out snmap create and update workflow
Have made sure that such errors are caught at NeutronVPN
level itself by separating out subnetmap create and update.
Update will not create subnetmap node with partial info, instead will
return if subnetmap node is absent.
Change-Id: I252d0164f123052d7090968325e214289b29daab
Signed-off-by: Abhinav Gupta <abhinav.gupta@ericsson.com>
epgoraj [Wed, 15 Mar 2017 05:07:59 +0000 (10:37 +0530)]
Adding some more debug logs to elan module
Change-Id: I6fe5f73c61367f32f59f67b9b49844f24e5cf388
Signed-off-by: epgoraj <p.govinda.rajulu@ericsson.com>
karthik [Sat, 18 Feb 2017 01:41:18 +0000 (07:11 +0530)]
Bug 7817 & Bug 7838: DHCP ARP flow is not added and irrelevant ARP flows
are installed in compute node.
* DHCP ARP flow is not added in Compute node
When VM's are created with multiple networks.
* VM's are created across networks, irrelevant ARP flows are
installed in compute node
* Fix: Compared with segmenationID and write the flows for table 20.
* This patch contains fix for both the Bug 7817 & Bug 7838.
Change-Id: Iab21d2fe391451dc1f6e9cc70db099105e990871
Signed-off-by: karthik <karthikeyen.p@hcl.com>
Koby Aizer [Thu, 2 Mar 2017 10:24:55 +0000 (12:24 +0200)]
Bug 7888: handle update of floating ip port
In Ocata, when a floating ip port is created, its deviceId may be
set to "PENDING", and an update will follow with the deviceId later on ([1])
This is causing both an uncaught exception in NeutronPortChangeListener.add()
which assumed deviceId is always a Uuid, and also there is no creation of
FloatingIpIdToPortMapping during update().
[1] https://review.openstack.org/#/c/396013/
Change-Id: Ide44ef016f0a8c576f02dfd28d9a4ee9d492841e
Signed-off-by: Koby Aizer <koby.aizer@hpe.com>
(cherry picked from commit
b8d44834604e071f050515b96a64d8e00a09c379)
Koby Aizer [Wed, 1 Mar 2017 10:02:16 +0000 (12:02 +0200)]
Bug 7878: provider interface MACs are installed on remote DPNs
MACs learnt on provider interfaces must not be installed on remote
DPNs.
This change includes 2 fixes:
1. handleInternalTunnelState event initiated installation on remote
DMAC flows for all elans in the remote DPN, without verifying that
this elan is not Flat/Vlan.
2. Protection on setupRemoteDmacFlow to make sure no other code flows
will install remote DMAC flows on MACs learnt on provider interfaces.
Change-Id: I5ae288ed525dc1ae2f5be1bec5215f2e8f01f8c5
Signed-off-by: Koby Aizer <koby.aizer@hpe.com>
Viji J [Fri, 10 Mar 2017 08:48:55 +0000 (14:18 +0530)]
Rectified incorrect help usage displayed for BGP add-neighbor cli command
odl:configure-bgp -op add-neighbor displays incorrect help.
Change-Id: Ic1e7bd747a07de0fdc562eb59d8bc835558900ac
Signed-off-by: Viji J <viji.j@ericsson.com>
Olga Schukin [Mon, 27 Feb 2017 14:19:22 +0000 (16:19 +0200)]
Bug 7787 - missing flows in T21
when VPN interface up before tunnel info is available in operational DS,
it causes missing flows in T21
Fix: always handle the add event inside TunnelInterfaceStateListener -
even for a tunnel which state is Down
Change-Id: I41c3efbeeb5740788f58ced57a698a73ef70b466
Signed-off-by: Olga Schukin <olga.schukin@hpe.com>
Signed-off-by: Sam Hague <shague@redhat.com>
Vivekanandan Narasimhan [Wed, 8 Mar 2017 09:06:16 +0000 (14:36 +0530)]
Bug 7931: SubnetRoue re-election to be triggered on disconnected nodes
Subnet Route reelection algorithm is triggered today only when
the anchor ports on the subnet go to DOWN state (or) if such
anchor ports on the subnet are being removed.
However, there are situations where a switch hosting the anchor
port can be disconnected on the openflow control-path and can
continue to remain so indefinitely.
Under such circumstance, new non-neutron IPs become undiscoverable
as the switch hosting the anchor port cannot punt discoverable IP
packets to the controller.
So this fix attempts to always trigger reelection of nodes for
subnetRoute hosting as soon as the current anchor node(or switch)
is detected to be disconnected from the controller.
Change-Id: Idd8be8e2f1940c4840d7d52726ac5187ed33d28f
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>
(cherry picked from commit
f9d156dc32aff937425eb586bee3ceefda069091)
ehvkand [Thu, 9 Mar 2017 06:09:11 +0000 (11:39 +0530)]
Bug 7876 : After router association to L3vpn, one of the VM ip is not
removed from router interface to BGPVPN
When a VPN associated to router is deleted and a same VPN is associated
back to same router , there is race condition , where interface is not found
in VpnInterfaceOp and remove event is being ignored.
Hence interface remains in router VPN.
It is also seen that , one of interface from same subnet is swapped and others are
not. Its due to race condition that exist in NeutronVpnManager where
create and update can race with each other. Currently create uses DJC ,
where as update doesn't.
Change-Id: I354c3ce93f965580c735f1153a043e2514d68af7
Signed-off-by: ehvkand <hanamantagoud.v.kandagal@ericsson.com>
Aswin Suryanarayanan [Mon, 6 Mar 2017 11:43:44 +0000 (17:13 +0530)]
Bug 7885 - CSIT Sporadic failures -
tempest.scenario.test_port_security_macspoofing_port
On port-security update bind/ubind service is now called.
Change-Id: I46cfae8b13a3f67647bfc501b3e55faf8c827a37
Signed-off-by: Aswin Suryanarayanan <asuryana@redhat.com>
(cherry picked from commit
9a9c988bfa3574126b38003d5a8eb713536809c2)
Shashidhar Raja [Wed, 22 Feb 2017 13:29:52 +0000 (18:59 +0530)]
Bug 7839: ACL: ACL flows are not deleted from source host during VM
migration
- In the current code, ACL flows are being added in Interface state
listener and deleted from Interface config listener. As during
migration, only interface state listener gets triggered, Config
Interface and Interface state listeners are re-formatted slightly as
explained below to fix this issue and also to support delete VM
without any problems.
Details of changes are as below:
(a) bind/unbind moved to Config interface listener
(b) All other ACL flows programming (add/delete) now being done from
Interface state listener
- Also, bind/unbind service code is updated to use
DataStoreJobCoordinator to avoid ConflictingModificationExceptions
Change-Id: I4cf0c0d58eecbceafbed740f0201e128c196f555
Signed-off-by: Shashidhar Raja <shashidharr@altencalsoftlabs.com>
Faseela K [Mon, 6 Mar 2017 06:20:26 +0000 (11:50 +0530)]
Use the right service name when binding service
Making ELAN and ACL bind services with the right service name
Change-Id: I022a0ff4aaff4772870413890ce8a9665af5a309
Signed-off-by: Faseela K <faseela.k@ericsson.com>
(cherry picked from commit
4e608f1f3050c1ccec402a800a04b9c3bf1403b5)
eamakol [Thu, 2 Mar 2017 05:55:33 +0000 (11:25 +0530)]
corrected the population of BGP Total Prefixes counter
Change-Id: I46106f3e7f96aef8d3c751f66db438b77bcd46d7
Signed-off-by: eamakol <amaranadh.kolla@ericsson.com>
Alon Kochba [Tue, 28 Feb 2017 15:52:43 +0000 (17:52 +0200)]
Bug 7856: Reverse SNAT flows order to minimize race possibility
Conflicts:
vpnservice/natservice/natservice-impl/src/main/java/org/opendaylight/netvirt/natservice/internal/ExternalRoutersListener.java
vpnservice/natservice/natservice-impl/src/main/java/org/opendaylight/netvirt/natservice/internal/NaptEventHandler.java
vpnservice/natservice/natservice-impl/src/main/java/org/opendaylight/netvirt/natservice/internal/NaptSwitchHA.java
Change-Id: I145905151348d2fa2670756f24e17c16d6239885
Signed-off-by: Alon Kochba <alonko@hpe.com>
ehvkand [Fri, 3 Mar 2017 08:22:13 +0000 (13:52 +0530)]
Bug 7714: VPN Operational Interfaces not getting removed at all.
This can be a potential problem for temporary CSIT failures that happens on
VPNService suite.
The VPNInterface is not getting removed at all permanently from
VpnInterface Op DS. When VM interface has secondary adj like extra route
or invisible ip , the order in which cleanupvpninterface
worker DJC is called not guaranteed.
The regression was caused by review:
https://git.opendaylight.org/gerrit/#/c/51224/
VPNInterfaces not deleted from Oper DS
Change-Id: I1b2a1d139030f707cec93eb5a5e6a690e010e336
Signed-off-by: ehvkand <hanamantagoud.v.kandagal@ericsson.com>
eupakir [Wed, 1 Mar 2017 10:00:01 +0000 (15:30 +0530)]
Bug 7831 : BgpRouter receives unnecessary events
Suppressed unwanted advertisement of routes, that were being advertised
due to a
breakage in the VPN Engine.
When BGP Client is not configured, route/vrf advertisement used to fail
and throw lot of exception. If BGP Client is not connected, the
process of dispatching routes/vrf to DC-GW via BGP Client should not
happen
at all.
Change-Id: I25b6abe7a67efb4621745933f869f7034eb9cd51
Signed-off-by: eupakir <kiran.n.upadhyaya@ericsson.com>
Marcos López Samaniego [Tue, 28 Feb 2017 16:23:51 +0000 (17:23 +0100)]
Bug 7881 - Traffic drops when not matching UL SC starting
in a VPNPseudoPort
When programming the L3VPN to SCF pipeline, a new flow is needed in
table 17
Change-Id: I31cf7817eb38b47e5c665c4f0362d0baa9554af4
Signed-off-by: Marcos López Samaniego <marcos.lopez.samaniego@ericsson.com>
karthikeyan [Mon, 27 Feb 2017 16:07:03 +0000 (21:37 +0530)]
Bug 7861: No ping response from FIP on 1st router when adding 2nd FIP
Problem Description:
====================
Table Miss Entry for SNAT/DNAT in FIB Table (table=21->26 flow) is
overriting when multiple router is present instead of creating new entry
for each VPN Instance.
Each router is associated with unique external network and each external
network is associated with unique internet BGP VPN Instance.
Example topology:
router-1/external-network-1/bgp-vpn-1
router-2/external-network-2/bgp-vpn-2
Both router-1 and router-2 are presented on single compute node. In this
scenario router-2 ping traffic only is working as expected. Since table 21
doesn't have the table miss entry for router-1 with bgp-vpn-1 instance.
Solution:
=========
In ExternalNetworkListener while creating table miss entry for SNAT/DNAT
in table 21, added "vpn-id" attribute also in creating flow reference as
same as implemented in the class
SNATDefaultRouteProgrammer.buildDefNATFlowEntity(). After this code change
in table 21 miss entry for SNAT/DNAT is created for each bgp-vpn instance.
Change-Id: Iac909e628f1a4134526e6df0b640149c8d45cd20
Signed-off-by: karthikeyan <karthikeyan.k@altencalsoftlabs.com>
epgoraj [Mon, 27 Feb 2017 07:35:15 +0000 (13:05 +0530)]
BUG:7824 ModifiedNodeDoesNotExistException
Change-Id: I32b64af56b5b7c5ed8d2af9ed3ca5b47b68a153f
Signed-off-by: epgoraj <p.govinda.rajulu@ericsson.com>
Alon Kochba [Mon, 20 Feb 2017 13:21:20 +0000 (15:21 +0200)]
Cleanup errors for networks of unsupported type
Reword and lower debug level of network configurations
for networks without a provider extension or with an
unsupported type.
Change-Id: I3b58a79b2fb72fe2c56328f79e0962e366a299ff
Signed-off-by: Alon Kochba <alonko@hpe.com>
Signed-off-by: Sam Hague <shague@redhat.com>
cgowdru [Mon, 13 Feb 2017 12:34:47 +0000 (18:04 +0530)]
Bug 7775: Using DJC for NAT Interface-state Listeners
Description : With mdsalutil is having a THREAD_POOL_SIZE=1, all the
add(), update(), and remove() methods handling the application logic were
on the same thread. This in turn was delaying all the interface-state
listeners.
Changes are done to make use of DataJobCordinator so that all the NAT
application logic get executed by a seperate thread so that the main
thread is freed-up for other listener execution.
Change-Id: I0c7f3f70715b01a2dcb54d251a36a2f9ff21357e
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>
Periyasamy Palanisamy [Thu, 23 Feb 2017 06:53:47 +0000 (12:23 +0530)]
Bug 7824 - ModifiedNodeDoesNotExistException
* added potential fix to guard against above exception
while handling elan interface remove/interface state remove
handling
Change-Id: Iad5145eafd9398718171880825f1684189405398
Signed-off-by: Periyasamy Palanisamy <periyasamy.palanisamy@ericsson.com>
Periyasamy Palanisamy [Mon, 13 Feb 2017 10:39:38 +0000 (16:09 +0530)]
releasing dcn thread once tunnel interface state dcn delivered
* This change is about releasing dataTreeChangeHandlerExecutor's thread
immediately after DCN is delivered to service module. This allows to
improve DataTreeChangeListener notification time when there are multiple
listerners listening to same data tree path.
* For ELAN, DJC is used to run tunnel interface state DCN related tasks in async
way using ElanUtils#getElanInterfaceJobKey
Change-Id: If3128e623ac3eae81fbec17320bf54d425b7f6f5
Signed-off-by: Periyasamy Palanisamy <periyasamy.palanisamy@ericsson.com>
cgowdru [Mon, 13 Feb 2017 08:07:38 +0000 (13:37 +0530)]
Bug 7780 : NAT RPC's for getting SNAT/DNAT translation information.
Description : With the Current DS structure, it's is bit difficult to
identify information about what is the current DNAT/SNAT configuration
done fo the given router-uuid/vpn-uuid.
It would be good if we have a RPC's which can provide a information as
below.
input- router-uuid/vpn-uuid
SNAT output - router-name, internal-ip/port, external-ip/port,protocol
DNAT output - router-name,internal-ip,floating-ip
Change-Id: I00b6576d2a752b53e94ad28c611a06794c65456c
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>
ehvkand [Thu, 23 Feb 2017 13:33:35 +0000 (19:03 +0530)]
Bug 7815: Using DJC for VpnManager Interface-state Listeners
With mdsalutil is having a THREAD_POOL_SIZE=1, all the add(), update(), and
remove() methods handling the application logic were on the same thread.
This in turn was delaying all the interface-state listeners. Changes are
done to make use of DataJobCordinator so that all the VpnManager
application logic get executed by a seperate thread so that the main
thread is freed-up for other listener execution.
Change-Id: I926e0742fdc7955450c9423ce5026f9177ce21bb
Signed-off-by: ehvkand <hanamantagoud.v.kandagal@ericsson.com>
Tali [Thu, 23 Feb 2017 13:35:25 +0000 (15:35 +0200)]
Bug 7843 - Missing buckets in ELAN BC group installation during OVS restart
L2 pipeline can contain malformed flows/groups if OVS restart occurs in parallel
due to oper DS temp delete during restart
Possible solution would be to check for config DS inventory node instead of oper DS
so flows would be restored with correct content after the OVS is up again
Change-Id: I5aee6c2af85bca1beb98dab52bfa9181975192b1
Signed-off-by: Tali <tali.ben-meir@hpe.com>
Akash [Mon, 13 Feb 2017 15:39:44 +0000 (21:09 +0530)]
Bug 7786 Delete and re add of access port handling
After deleting access port in l2gw device and upon readding , vlan
bindings were missing .
Fix is to resync upon such case from config topology.
Change-Id: I993616f6ea90312af2ae5830e444c1918f8b48bb
Signed-off-by: Akash <a.k.sahu@ericsson.com>
Miguel Perez [Mon, 16 Jan 2017 17:30:16 +0000 (18:30 +0100)]
Bug 7772 - Service Chaining is not being applied to VMs in the L3VPN
+ Problem: Ping from outside DC-GW to a VM located inside DC-GW
works well in the Uplink, that is, packets going from outside
DC-GW towards the VM are hijacked by SCF Pipeline and delivered
to VM. However, when VM responds (Downlink), the packets are sent
to DC-GW right away via L3VPN pipeline. Service Chaining is not
applied for the return path.
+ Solution: try to divert to SCF Pipeline the packets that come from
the VM too, not only the packets coming from the external Tunnel
port.
+ How: by adding a listener on VpnInterface on CONFIG DS.
When a VpnInterface is added, the listener checks if its VPN is
involved in a ServiceChain or not. If so, the interface is bound
to SCF too.
Change-Id: Id1051bc8ec180589583a76524c0b2d456ba7b85a
Signed-off-by: Marcos López Samaniego <marcos.lopez.samaniego@ericsson.com>
eupakir [Tue, 21 Feb 2017 15:35:21 +0000 (21:05 +0530)]
Adding debug statements to track caching of Operational Vpn Instances
Conflicts:
vpnservice/vpnmanager/vpnmanager-impl/src/main/java/org/opendaylight/netvirt/vpnmanager/cache/listeners/VpnOpInstanceCacheManager.java
Change-Id: I0bb17dec5714e8f68085bf1b70e2b2fa1f2f16ae
Signed-off-by: eupakir <kiran.n.upadhyaya@ericsson.com>
Miguel Perez [Wed, 15 Feb 2017 14:08:55 +0000 (15:08 +0100)]
Fix priority in IntervpnLink flows installed in LFIB
+ InterVpnLink flows installed in LFIB must have higher priority than
regular ones (which is typically 1 or 10)
Change-Id: Ifb6a6e8d91f83b65ffda356cd16ad878a1ae11ee
Signed-off-by: Miguel Perez <francisco.miguel.perez@ericsson.com>
epgoraj [Mon, 20 Feb 2017 12:45:53 +0000 (18:15 +0530)]
adding lport tag for temporary mac learning
* loading lport tag into NxmNxReg1 while binding elan service
to the interface.
* adding NxmNxReg1 match for table 49 using relearn action
Change-Id: I2914c08157bf33fdb033b7656f1692b6be3ea690
Signed-off-by: Periyasamy Palanisamy <periyasamy.palanisamy@ericsson.com>
Signed-off-by: epgoraj <p.govinda.rajulu@ericsson.com>
Alon Kochba [Sun, 12 Feb 2017 16:29:05 +0000 (18:29 +0200)]
Fix several NPEs showing up in CSIT
1. ElanUtils - an ELAN instance can have no segments
2. TEPChangeListener - TEP could be missing in interfaces-state
3. VpnInterfaceManager - gateway IP/MAC could be missing when creating
ARP responder flows.
4. BgpConfigurationManager - NextHop in list could be null
5. TransportZoneNoticationUtil - not all interfaces have a neutron port
Change-Id: Ibbf6e5834cd01389e29e36736073927cdd3dbb1a
Signed-off-by: Alon Kochba <alonko@hpe.com>
Signed-off-by: Sam Hague <shague@redhat.com>
Tali [Wed, 15 Feb 2017 16:26:15 +0000 (18:26 +0200)]
Fix BFD regression
When tunnel and elan-interface goes up in parallel, there is a race in writing
DMAC flows. The result could be DMAC flow with drop action for the affected interface
Change-Id: I080e222025d59f0b482cd6c07563c168646afc76
Signed-off-by: Tali <tali.ben-meir@hpe.com>
(cherry picked from commit
19c5da8fc44c9435dcd597c701eaccdba9e1b595)
Suraj Ranjan [Tue, 14 Feb 2017 08:36:57 +0000 (14:06 +0530)]
BUG7748: Subnet-op-data empty after cluster reboot
SubnetAddToVpn notification is fired from add() of SubnetMapListener
also instead of firing from update() only.The reason for doing this is
after reboot this event comes in subnetmap add().
Change-Id: I247753ba1a52e896a4bd2ec7f96c3b5efae35055
Signed-off-by: Suraj Ranjan <suraj.ranjan@ericsson.com>
Tali [Thu, 16 Feb 2017 13:27:24 +0000 (15:27 +0200)]
Bug 7790 - Attempting to install RNH on local DPN for FIB with custom instructions
On VxLAN network there is no severe side affect other than trashing the log
For VLAN/flat networks this is a major issue since VrfEntryListener is not able to
find any local DPN hence all FIB entries are considered RNH and the actual local
FIB entry created using FibRpcService.createFibEntry() will be overridden by the
faulty RNH on the local DPN.
Change-Id: I49f6660608337fa9baf178179d589d9c6980964e
Signed-off-by: Tali <tali.ben-meir@hpe.com>
Alon Kochba [Mon, 6 Feb 2017 19:24:16 +0000 (21:24 +0200)]
Use Objects equals instead of == where necessary
This might actually fix bugs and change behavior,
there were places where Java objects were compared
using == instead of equals(), this patch corrects
those cases.
Conflicts:
vpnservice/cloud-servicechain/cloud-servicechain-impl/src/main/java/org/opendaylight/netvirt/cloudservicechain/listeners/VrfListener.java
vpnservice/cloud-servicechain/cloud-servicechain-impl/src/test/java/org/opendaylight/netvirt/cloudservicechain/matchers/FlowMatcher.java
vpnservice/neutronvpn/neutronvpn-api/src/main/java/org/opendaylight/netvirt/neutronvpn/api/utils/NeutronUtils.java
vpnservice/neutronvpn/neutronvpn-impl/src/main/java/org/opendaylight/netvirt/neutronvpn/NeutronvpnManager.java
vpnservice/vpnmanager/vpnmanager-impl/src/main/java/org/opendaylight/netvirt/vpnmanager/ArpReplyOrRequest.java
vpnservice/vpnmanager/vpnmanager-impl/src/main/java/org/opendaylight/netvirt/vpnmanager/VpnSubnetRouteHandler.java
Change-Id: I6996e168966fb660d62cca3e4c81635a6c1f6b06
Signed-off-by: Alon Kochba <alonko@hpe.com>
Alon Kochba [Wed, 15 Feb 2017 13:28:30 +0000 (15:28 +0200)]
Update netvirt guide - correct DB DROP procedure
Change-Id: I99d19c118d05ade5a9c77f6a98f75abb1b8c7ddc
Signed-off-by: Alon Kochba <alonko@hpe.com>
K.V Suneelu Verma [Sun, 5 Feb 2017 09:14:56 +0000 (14:44 +0530)]
bug 7599 added l2gw validate cli
Added a helper cli to validate various config and operational
data models.
Checks if the config and operational datastore differ
Checks if the ha parent node and child nodes differ
Checks if the logical switches/ mcast macs and vlan bindings
are created for the configured l2gw connection
Change-Id: I96924a520d30985b98508afdaccbe379ef54d76e
Signed-off-by: K.V Suneelu Verma <k.v.suneelu.verma@ericsson.com>
K.V Suneelu Verma [Tue, 3 Jan 2017 06:43:10 +0000 (12:13 +0530)]
bug 6589 l2gw cluster reboot fixes
registerd the group listener
created the l2gw cache upfront
fixed npe in terminationPoint command
Change-Id: I5e4e4d68e650a2c68d3c7b4bf464dc29815e0d48
Signed-off-by: K.V Suneelu Verma <k.v.suneelu.verma@ericsson.com>
VinothB [Mon, 13 Feb 2017 04:22:04 +0000 (09:52 +0530)]
Bug 7606: Fix for missed tunnel flows, after VM live migration
* Skip deleting the flows in table 110 and table 20 when processing an
interface deletion event from OVSDB in the case that the interface
deletion is due to VM migration.
* Remove the Port Name from flowName so that, the same entry gets
updated during the update notification processing while live migration.
Change-Id: I2b917e50e07af4ed840d811a9c3f0a131281bd12
Signed-off-by: VinothB <vinothb@hcl.com>
Hideyuki Tai [Thu, 26 Jan 2017 01:56:17 +0000 (01:56 +0000)]
Bug 7773: Objects should be compared with "equals()".
Additional changes:
* Removed unnecessary repetitive calls.
Change-Id: I607ce073ee9ec4289c80dccd96a4b5222e4d9772
Signed-off-by: Hideyuki Tai <Hideyuki.Tai@necam.com>
Hideyuki Tai [Wed, 25 Jan 2017 21:43:00 +0000 (21:43 +0000)]
Improve log messages.
* Correct wrong messages.
* Change the log level of some messages into the appropriate level.
Change-Id: I3e238611ebc9c9fc42aa800caa2a44621880c9aa
Signed-off-by: Hideyuki Tai <Hideyuki.Tai@necam.com>
Sam Hague [Thu, 9 Feb 2017 19:27:51 +0000 (19:27 +0000)]
Revert "BUG7748: Subnet-op-data empty after cluster reboot"
This reverts commit
079b071c122cc3918b419d0ac62e504a3d89030a.
Change-Id: I0bce2abcb79aa43a85b12d260de525df6466f78a
Author: Sam Hague <shague@redhat.com>
Signed-off-by: Sam Hague <shague@redhat.com>
hari.pr [Fri, 3 Feb 2017 18:07:56 +0000 (23:37 +0530)]
Bug 7392: L2 Forwarding Table=110 Flows Missing
* Problem: While creation of very first subnet or Vm in CSIT environment,
respective 110 flow is missing in openstack nodes.
* Solution: Installed missing table#110 flows as part of Vxlan interface request.
Change-Id: Ie7eb35b50aaaafa79fd8ed0f0893ab37156f431e
Signed-off-by: hari.pr <hari.pr@hcl.com>
Vinh Nguyen [Sat, 4 Feb 2017 01:30:47 +0000 (17:30 -0800)]
BUG 7729: Remove redundant tunnel drop flow in table 110
- Remove methods: L2ForwardingService::programLocalTableMiss
and L2ForwardingService::programTunnelMiss. Both methods
generate the same tunnel drop flows in table 110.
The drop flows for individual tunnels are not needed
since there is a default drop flow in the same table.
- Remove the method L2ForwardingService::programVlanFloodOut
which is not used.
Change-Id: I0fb826094b6c03e58fb89c95600e7b84bc2c774f
Signed-off-by: Vinh Nguyen <vinh.nguyen@hcl.com>
Suraj Ranjan [Mon, 6 Feb 2017 16:06:37 +0000 (21:36 +0530)]
BUG7748: Subnet-op-data empty after cluster reboot
SubnetAddToVpn notification is fired from add() of SubnetMapListener
also instead of firing from update() only.The reason for doing this is
after reboot this event comes in subnetmap add().
Change-Id: I766e3836cbe8b0eafbf5fbb52791ca96eb6ccfa9
Signed-off-by: Suraj Ranjan <suraj.ranjan@ericsson.com>
Miguel Perez [Wed, 8 Feb 2017 17:24:13 +0000 (18:24 +0100)]
Bug 7680: Fix Nexthop when advertising to DCGW
+ for static routes whose nexthop is a VM, the leaked route to the other
VPN was being advertised to DCGW with the DPN where VM is located as
the nexthop of the route. However, the nexthop must be the DPN where
the InterVpnLink has been placed.
Change-Id: I1097e9c992c66053f499807fd411334a7113b509
Signed-off-by: Miguel Perez <francisco.miguel.perez@ericsson.com>
Miguel Perez [Tue, 24 Jan 2017 17:01:29 +0000 (18:01 +0100)]
Bug 7680: leaked routes not advertised to DC-GW
+ It's been found that the routes that are leaked to DC-GW
at InterVpnLink creation time (well, at InterVpnLink added
callback) are correctly advertised to DC-GW. However, the
routes that are leaked after the InterVpnLink has been
created are not being advertised to DC-GW.
+ These second type of leaked routes are handled on fibmanager,
but fibmanager does not have access to BgpManager. So the
solution consists in moving part of the logic from fibmanager
to vpnmanager where the BgpManager is accessible.
+ InterVpnLinkService class, which was just an internal class in
vpnmanager-impl, is renamed to InterVpnLinkLocator whose only
responsibility is to search for the most suitable DPNs to
place new InterVpnLinks. The name change is to avoid confusion
with IVpnLinkServiceImpl which is the implementation of the
public interface 'IVpnLinkServce'.
+ Besides the renaming of InterVpnLinkService, most of its code
is moved to IVpnLinkServiceImpl, and that's why this change
may llok big.
+ Note: there is a <wip> [1] for taking the route leaking out
of the fibmanager completely.
[1] https://git.opendaylight.org/gerrit/#/c/48599/
Change-Id: I1a7311345307716ef223c707b22c6ecec66d97a2
Signed-off-by: Miguel Perez <francisco.miguel.perez@ericsson.com>
Tali [Thu, 15 Dec 2016 09:15:07 +0000 (11:15 +0200)]
Bug 7372 - Supress error of NAPT switch selection failure before router-dpn association
NAPT switch selection will take place once the router-dpn association is made
All missing NAPT/non-NAPT flows will be recovered as the result of selection
This review fixes only the log levels.
There can still be some missing NAPT flows in this scenario
The resolution for this is still WIP and covered by
https://bugs.opendaylight.org/show_bug.cgi?id=6866
Change-Id: Ib655c1fbf1c43f29a75223a0b48d410d47ef75d9
Signed-off-by: Tali <tali.ben-meir@hpe.com>
karthik.prasad [Thu, 12 Jan 2017 15:27:25 +0000 (20:57 +0530)]
Setup SMAC on routed packets destined to virtual endpoints
All L3 Routed packets destined to virtual endpoints in the datacenter
managed by ODL do not carry a proper source-mac address in such frames
put out to virtual endpoints. With the fix all the packets destined to
virtual endpoints shall have gateway mac as source mac. If gateway mac
is not recoverable for the subnet on which the virtual endpoint resides,
then the connected mac address of that virtual endpoint will be used as
the source mac address. Here the connected mac address represents the
mac-address on the tapxxx (or) the vhuxxx port on OVS Datapath for that
virtual endpoint.
VrfEntry yang model is used to represent the gateway source mac.
Sample Group Entry:
group_id=150001,type=all,bucket=actions=set_field:fa:16:3e:a1:4b:61->eth_src,set_field:fa:16:3e:21:b8:d3->eth_dst,load:0x300->NXM_NX_REG6[],resubmit(,220)
Spec Reference link:
https://git.opendaylight.org/gerrit/#/c/49807
Conflicts:
vpnservice/fibmanager/fibmanager-impl/src/main/java/org/opendaylight/netvirt/fibmanager/NexthopManager.java
vpnservice/vpnmanager/vpnmanager-impl/src/main/java/org/opendaylight/netvirt/vpnmanager/VpnInterfaceManager.java
vpnservice/vpnmanager/vpnmanager-impl/src/main/java/org/opendaylight/netvirt/vpnmanager/VpnUtil.java
Change-Id: I2e75215fd75f051135d702ab63cb792b6cf4cb55
Signed-off-by: karthik.prasad <karthik.p@altencalsoftlabs.com>
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>
Signed-off-by: Achuth Maniyedath <achuth.m@altencalsoftlabs.com>
karthikeyen.p [Wed, 18 Jan 2017 04:10:17 +0000 (09:40 +0530)]
Bug 7445: Improve the performance on bulk create.
* Reduced read calls when SG associated with the VMs.
* We have used cache mechanism to store neutron port information
that was read from DS.
Change-Id: I3b243ddf3656e48392593c6fc1086713d1fd1af3
Signed-off-by: karthikeyen.p <karthikeyen.p@hcl.com>
Abhinav Gupta [Thu, 12 Jan 2017 11:10:02 +0000 (16:40 +0530)]
Bug 7733: NeutronVPN: Error out if VxLAN/VLAN network configured without seg-id
This adds an error log based on the presence of segmentation-id when VxLAN
or VLAN based neutron networks are configured via Neutron.
Additionally, some cleanup done:
1. Moved some impl specific utils to NeutronVpnUtils
2. Moved constants to NeutronConstants
3. Renamed SubnetMapListener class to SubnetMapChangeListener
4. Removed unused xml dependency from QosListener
Change-Id: I2d22773e4e7bdfa927c01d679542fa36146e625a
Signed-off-by: Abhinav Gupta <abhinav.gupta@ericsson.com>
Olga Schukin [Tue, 31 Jan 2017 10:35:55 +0000 (12:35 +0200)]
Bug 7714 - Vpn Interface not deleted from oper DS
sometimes Vpn Interface not deleted from operational/l3vpn:vpn-interfaces model,
if there is more than one adjacency to this vpn interface.
Change-Id: I45e4ca20db6c5aa9f5c280fea8726a5ed8aa654b
Signed-off-by: Olga Schukin <olga.schukin@hpe.com>
Abhinav Gupta [Wed, 1 Feb 2017 13:09:40 +0000 (18:39 +0530)]
Bug 7720: create/delete VPN CLI handling addition/removal of subnets
Following changes have been done as part of the bug fix:
1. Addition of subnets to VPN done via associateNetworks RPC
2. Deletion of subnets from VPN done via dissociateNetworks RPC
3. (1) is done only if createVPN succeeds
4. Deletion of VPN is done only if (2) succeeds
Change-Id: I21d25afa6b3ba56dae56abe2536afcfe5f0476b4
Signed-off-by: Abhinav Gupta <abhinav.gupta@ericsson.com>
Sridhar Gaddam [Thu, 19 Jan 2017 14:45:14 +0000 (20:15 +0530)]
Bug 7601: Cleanup Elan instances when a network is deleted
The following patch in Neutron[1] modifies the Network delete sequence.
Prior to a network delete event, we now see an update event that removes
the networkType. So, in delete event, when we try to check for the
networkType, it would be empty and there is an exception. Because of this,
elanInstances are not getting deleted. In this patch, we avoid the check
for networkType since the network itself would not be created in the first
place if its not a supported networkType.
[1] https://review.openstack.org/#/c/317358/
Change-Id: I349a51f4df666ef2c2912ee552f159b6af16de76
Signed-off-by: Sridhar Gaddam <sgaddam@redhat.com>
Vinh Nguyen [Wed, 1 Feb 2017 01:06:50 +0000 (17:06 -0800)]
BUG 7717 Fix OOM when defining large number of networks
Use NodeId instead of Node as key for the cache
ProviderNetworkManagerImpl::nodeToProviderMapping
The size bridge Node object is huge when there are
large number of network/subnets on the bridge.
Change-Id: I6956318477dfa524ce20680d12bba3c7818540ec
Signed-off-by: Vinh Nguyen <vinh.nguyen@hcl.com>
Vishal Thapar [Mon, 30 Jan 2017 07:08:12 +0000 (12:38 +0530)]
BUG 7488 Add option to disable auto bridge creation
This adds a new param to elan-config, auto-create-bridge that defaults to
true. When false, netvirt will neither try to create br-int if it doesn't
exist, nor add any configuration if it exists. This is to be used in
scenarios where br-int is already created and configured.
Change-Id: I39c2770a0c0351d9e1b1fecd2642f473650e6c52
Signed-off-by: Vishal Thapar <vishal.thapar@ericsson.com>
(cherry picked from commit
1dd6d15f82b1f8f4b9897cca5c8ca71f6a49f898)
Vishal Thapar [Thu, 19 Jan 2017 13:21:11 +0000 (18:51 +0530)]
BUG 7591: Allow configuration of inactivity_probe and max_backoff for OVS
Change-Id: I7881934df38c32cd40e156944c3ee46804d5a4e4
Signed-off-by: Vishal Thapar <vishal.thapar@ericsson.com>
karthikeyan [Wed, 25 Jan 2017 08:05:05 +0000 (13:35 +0530)]
Bug 7667: SNAT table 46&44 are not getting programmed when private BGPVPN
Problem Description:
=====================
SNAT: Table 46(OUTBOUND_NAPT) and table 44 (INBOUND_NAPT) are not getting
programmed (SNAT Forward Traffic and SNAT Reverse Traffic) with NAT
translation when router is
associated with Private/Internal BGP-VPN.Due to this issue, SNAT traffic
is getting failed when interval bgp-vpn is configured.
Root Cause of the Problem:
==========================
As part of bug (Bug 7478) fix SNAT traffic is used to send router-gw mac
address instead of VM mac. With this code changes router gw-mac was
getting from external-router Yang Model using router name as key.
When private/internal bgp-vpn is configured router-id (Internal Private
VPN ID) should be replaced with created private/internal bgp-vpn id. With
this approach when querying external-router Yang model for getting
router-gw mac address using new router-id (bgp-vpn id) will not match with
router-name.
Solution:
===========
When querying external-router yang model for getting router-gw mac address
it should use old router-id instead of bgp-vpn id. For doing this change
in NAPTEventHandler re-aligned the router-gw mac address code to
appropriate place to be working as expected.
Change-Id: Ieefd799d6b2349464bfb737e3fcc8101036c3619
Signed-off-by: karthikeyan <karthikeyan.k@altencalsoftlabs.com>
Vishal Thapar [Sun, 22 Jan 2017 09:30:08 +0000 (15:00 +0530)]
BUG 7489: Add startup config file for elanmanager-config
This implements issue 2 mentioned in BUG 7489 by adding a new
configuration file in which user can specify default values for
elan-config at startup.
Note: Update-Strategy is set to None which means any runtime changes to
config will not be reflect till next restart
Path: etc/opendaylight/datastore/initial/config/
Config file name: netvirt-elanmanager-config.xml
Change-Id: I7e3c34166ffd3d8c72f860e1ea34a2c3cc79c309
Signed-off-by: Vishal Thapar <vishal.thapar@ericsson.com>
Abhinav Gupta [Mon, 30 Jan 2017 07:51:10 +0000 (13:21 +0530)]
Bug 7700: create-l3vpn (REST/CLI) should not allow another VPN to use the same VPNID
Change-Id: I4481db51b17e92b2df7afab18b41bcb4e1a23df0
Signed-off-by: Abhinav Gupta <abhinav.gupta@ericsson.com>
Yakir Dorani [Sun, 22 Jan 2017 17:15:09 +0000 (19:15 +0200)]
Fix ElanStatusMonitorJMX failing upon bundle reinitialization.
Change-Id: I9e04e3ee1b9edc6000439e18b35c729f9ec0f92b
Signed-off-by: Yakir Dorani <yakir.dorani@hpe.com>
(cherry picked from commit
7f1f03e93f02fddd3002066d87b52ad782566c53)
Olga Schukin [Thu, 26 Jan 2017 06:45:16 +0000 (08:45 +0200)]
SNAT tests failures
The rule with empty action list was installed to T47
Change-Id: I8fb81cafb2a7bfd392ce24b73dedf14ab29608bb
Signed-off-by: Olga Schukin <olga.schukin@hpe.com>
David Goldberg [Thu, 19 Jan 2017 16:12:38 +0000 (18:12 +0200)]
BUG7308: fix leaf to leaf traffic
Change-Id: Iabae4f9fb7a2665f41f705ad0a6516a2d96ec77b
Signed-off-by: David Goldberg <gdavid@hpe.com>
(cherry picked from commit
adb975a4ca8acc029e6db3d8f7f558669f375748)
Bertrand [Thu, 26 Jan 2017 03:19:00 +0000 (19:19 -0800)]
BUG-7461
Align Legacy Netvirt with New Netvirt's design of not
removing integration bridge from the config DS.
Change-Id: I9f5da8344747f854828cddedd8c91f633b7b666f
Signed-off-by: Bertrand <bertrand.low@hcl.com>
Victor Pickard [Tue, 24 Jan 2017 12:10:09 +0000 (07:10 -0500)]
Bug 7669: Add multi-provider network support to NetVirt for L2 Gateway.
Multi-provider network support is needed to be able to support
connecting VLAN based networks to VXLAN based networks,
using L2GW/HWVTEP.
Slides describing the network, and use cases, can be
found here:
https://docs.google.com/presentation/d/10Mg3OLkqXiEOcY0psB4aI-TaHRvOX0NWAOlfkmtN6RY/edit#slide=id.g1b8ae7086a_0_402
Testing
-------
Setup and successfully tested use cases 1-3, as described in
slides above.
Note: There is a corresponding Nova bug that must be fixed
for this to work. Basically, Nova needs to add support
for multi-provider networks. This work is in progress.
Change-Id: Iba4e52210e8c27e0ec35fe68b1fba3970b3a66c8
Signed-off-by: Victor Pickard <vpickard@redhat.com>
Alon Kochba [Wed, 25 Jan 2017 16:47:22 +0000 (18:47 +0200)]
Lower debug level when truncating provider port name
This print shows up a lot in the log, as it is
called on multiple events, whenever the configured
provider port name is longer than 15 chars.
Lowering debug level as it is mostly an internal
translation.
Change-Id: Ic317c96c66cd876e2fd3628b4985ea0abb6d5152
Signed-off-by: Alon Kochba <alonko@hpe.com>
Maxime Millette-Coulombe [Thu, 19 Jan 2017 21:01:41 +0000 (16:01 -0500)]
[BUG 7543] Replace the request function used by restangular
Restangular as two methods to fetch data: get and getList.
Those two methods was doing the same things, but it seems
that getList now expect a List of JSON and not an object
who caused the application to crash.
Change-Id: I4aa51860cd03f3f6a04d2322518c7c9c7a1b3092
Signed-off-by: Maxime Millette-Coulombe <mmcoulombe@inocybe.com>
(cherry picked from commit
55893ff64b7c3fe7e7e50410a04ab57bdf1855b3)
Olga Schukin [Mon, 23 Jan 2017 14:52:06 +0000 (16:52 +0200)]
Bug 7660 Infinite loop while vpn instance removal
Sometimes VpnInstanceListener gets into endless loop because not all vpn
interfaces removed successfully.
Analysis:
The current interface counter value inside the function waitForOpRemoval()
can be less that original counter, but still not zero, then we are back to
beginning of the cycle without any action and sleep again.
The infinite loop doesn't allow to DataStoreJobCoordinator to run any
other job within same key.
Change-Id: I31ce3d38fa49212b7736888664a2d597d09f9666
Signed-off-by: Olga Schukin <olga.schukin@hpe.com>
Abhinav Gupta [Wed, 25 Jan 2017 18:49:30 +0000 (00:19 +0530)]
Bug 7384: CSIT Exception: NPE in deleteVpnInterface
getNeutronPort is returning null, causing port.getUUid() to result
in an NPE.
Further analysis necessary as to how it happened in the first place
since port delete event for the port in concern hasn't come.
The introduced check should ensure avoiding this NPE for now.
Change-Id: Ia4298d3b9f2b57b1a94805bca3ae34dffa2df4fe
Signed-off-by: Abhinav Gupta <abhinav.gupta@ericsson.com>
Olga Schukin [Thu, 19 Jan 2017 11:12:40 +0000 (13:12 +0200)]
Bug 7532 - arp responder rule sometimes missing after vm reboot
VRF entry is not removed during the port down because of wrong value (null string) in the next hop address list.
Change-Id: Ib0a0a84358cce116bdb9c4768ff9d9608f5c638b
Signed-off-by: Olga Schukin <olga.schukin@hpe.com>
Sam Hague [Sat, 14 Jan 2017 13:54:14 +0000 (08:54 -0500)]
Bug 7601 - Cleanup Elan instances when a network is deleted
Fix npe for getNetworkType() being null
Change-Id: I0e7ede43cdd7553b4525f5921b487f7e17c25b52
Signed-off-by: Sam Hague <shague@redhat.com>
Tomer [Sun, 25 Dec 2016 12:31:02 +0000 (14:31 +0200)]
Bug 7436: Handle VpnInterfaces of VpnInstance
The problem:
VpnOpDataSyncer.waitForVpnDataReady is using java wait method
that does not return the thread to the pool while waiting.
If there are more compute nodes then CPU threads, all threads
in DjC will wait and there will be not availabe thread to
execute the code which waitForVpnDataReady waits for.
The solution:
VpnInterfaceManager holds all vpn interfaces that cannot be processed till the
relevant vpn instance in ready.
Updated the synchronized lock blocks to user vpnInstanceName instead of this.
Added cleanup of removed interface
Change-Id: I66da144356e17e778f1d04e5720befbece565fd5
Signed-off-by: Tomer <tomer.pearl@hpe.com>
Signed-off-by: Koby Aizer <koby.aizer@hpe.com>
Miguel Perez [Fri, 13 Jan 2017 13:13:50 +0000 (14:13 +0100)]
Bug 7536: Static routes not handled when ivpnlink becomes active
+ Corner test case found that if the InterVpnLink is created when
there's no available DPN to be installed in, the InterVpnLink is
correctly created in ERROR state. Next step is to add some static
routes to the corresponding Neutron Router. Third step is to connect
some OVS node to the controller. At this point, the InterVpnLink
is correctly installed on that DPN and the State is correctly set
to ACTIVE... but the Static routes are forgoten.
+ It may look big but it is mostly moving code from
InterVpnLinkListener to InterVpnLinkService
Change-Id: I87018e6cba48146c12fbad0d506445931275977e
Signed-off-by: Miguel Perez <francisco.miguel.perez@ericsson.com>
Guy Sela [Thu, 12 Jan 2017 14:50:05 +0000 (16:50 +0200)]
Bug 7530 : ElanPacketInHandler mutexes are too coarse
Mutex for manipulating flows related to MAC was locking on
ELAN + MAC, when could lock ELAN + MAC + DPID. By locking
on the finer-grained mutex, we can support much bigger scale.
Problem occured in scale testing when 180 computes were
getting packet ins with the same MAC (External Network), and
every OVS was competing with all other OVSs for the lock of ELAN+MAC.
ElanPacketInHandler also updates a global data structure
of MACs that must be locked with the coarser mutex of
ELAN+MAC. This is why this patch splits the ElanPacketInHandler
into 2 jobs instead of one.
The transactions there also caused OptimisticLockException because
delete was called and submitted and then put was called and
submitted. This can't work due to the asynchronoush nature of
MD-SAL. There is no need in delete anyway if you call put, so
the delete was removed.
The 10 seconds protection was removed because it is unnecessary
as it already exists in the OVS with a learn rule.
The MAC migration code never worked because the condition could
ever be true there (it's a bug). I preserved the same behavior.
Fixed potential NPE in Etree code
Change-Id: I1aa038f9e0d271b94ffee6076654571b2e811d37
Signed-off-by: Guy Sela <guy.sela@hpe.com>
(cherry picked from commit
a30c620a4d5d3f2c4bc63fff433e691f6501ebf9)
Kency Kurian [Wed, 11 Jan 2017 19:19:34 +0000 (00:49 +0530)]
Bug 7533 : Fix for bind/unbind in DHCP service
- If the unbind is called from interface state listeners there could be
chances of ending up in a race condition in service bind logic and hence
the flow does not get removed.
- Moved the bind service call to interface config listener rather than state
so that the service flows get programmed as soon as the interface is UP.
Change-Id: Iadd3743e47efed9f256612fe8bab7085d3de2d54
Signed-off-by: Kency Kurian <kency.kurian@ericsson.com>
Vivekanandan Narasimhan [Tue, 17 Jan 2017 17:02:15 +0000 (22:32 +0530)]
Bug 7567: External subnet group is not updated with external gwmac
The external subnet group is not getting updated even though
the resolution for external gateway macaddress is successful.
The fix here is to make the the external subnet group programmer
to listen for Learned IP-MAC datastore which would really hold
the external-subnet-gateway-ip TO external-subnet-gateway-mac
mapping.
Also multiple external subnets can share the same subnet ip addresses
and so this patch fixes that too..
Change-Id: Id8c2ba8ace8dbeaf20ceb35d81f43a07e50c3a9f
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>
Guy Sela [Wed, 11 Jan 2017 15:30:40 +0000 (17:30 +0200)]
Bug 7528 : Don't learn the DMAC flows from other DPNs
When working with an elan of type VLAN or FLAT, we shouldn't
learn the known macs from other DPNs. Before calling
programRemoteDmacFlow, validate that the elan is not VLAN or FLAT.
Change-Id: Ib96546aff94bc18f830d72c1735c364f1a21156b
Signed-off-by: Guy Sela <guys.sela@hpe.com>
Marcos López Samaniego [Mon, 9 Jan 2017 14:35:15 +0000 (15:35 +0100)]
Bug 7525 - Inter-VPN link static/connected routes leaking not working
+ Enable static/connected routes in inter-VPN link creation.
+ Fixed some issues related to routes provisioned after the inter-VPN link.
These issues were affecting the leaking of every kind of route.
Change-Id: Ibb73f86f163f6fe7adf4a26b81ab1ead93855561
Signed-off-by: Marcos López Samaniego <marcos.lopez.samaniego@ericsson.com>
ehvkand [Mon, 16 Jan 2017 09:11:20 +0000 (14:41 +0530)]
Bug 7547 : Ping from DC-GW to invisible ip configured in VM is failing
Change-Id: Iba7869b3ccbaf3f5921c9f705628444aca630996
Signed-off-by: ehvkand <hanamantagoud.v.kandagal@ericsson.com>
Olga Schukin [Tue, 10 Jan 2017 08:11:00 +0000 (10:11 +0200)]
Bug 7497 - NAPT rules missed for second DPN
Install the group using synced mdsal write to improve the timing.
Change-Id: I3270fa1f1bc3bbccf11d1b02befe76db7f70fcc2
Signed-off-by: Olga Schukin <olga.schukin@hpe.com>
Sam Hague [Sat, 14 Jan 2017 13:23:46 +0000 (08:23 -0500)]
Fix links to openstack images
Change-Id: Ic361aa3b2530f9770c1b2ad6ab27b2760c8d265e
Signed-off-by: Sam Hague <shague@redhat.com>
karthikeyan k [Thu, 17 Nov 2016 08:19:09 +0000 (13:49 +0530)]
Bug 7478 : SNAT traffic to use router GW MAC
Problem Statement :
Currently SNAT traffic uses VM MAC as source MAC instead of Router Gateway
MAC
Solution:
Changes has been done for SNAT traffic as source MAC address as Router Gateway MAC Address in
NAPT_OUTBOUND_TABLE(Table=46)
Change-Id: I1563cc9f04d8e479c9c1173644a168262b448d80
Signed-off-by: karthikeyan k <karthikeyan.k@altencalsoftlabs.com>
Signed-off-by: karthikeyan <karthikeyan.k@altencalsoftlabs.com>
Achuth Maniyedath [Mon, 26 Dec 2016 13:26:31 +0000 (18:56 +0530)]
Spec to setup SMAC on routed packets destined to virtual endpoints
Change-Id: I72a2c008464ac99d7bd2796fae2d89030ee3826c
Signed-off-by: Achuth Maniyedath <achuth.m@altencalsoftlabs.com>
Signed-off-by: Sam Hague <shague@redhat.com>
Sam Hague [Thu, 12 Jan 2017 13:18:31 +0000 (08:18 -0500)]
Minor updates to openstack doc
Change-Id: I5097f2d911597c2c97b1564f62d0ba5a08538a5b
Signed-off-by: Sam Hague <shague@redhat.com>
Miguel Perez [Mon, 19 Dec 2016 15:07:16 +0000 (16:07 +0100)]
Bug 7405: IVpnLink routes not removed from BGP on cascade
+ When the InterVpnLink is removed, the routes that had been
advertised to BGP are not withdrawn in consequence
Change-Id: I795d1ebef26428561eb7307e5a84efdc50343cb7
Signed-off-by: Miguel Perez <francisco.miguel.perez@ericsson.com>
David Goldberg [Mon, 9 Jan 2017 16:44:23 +0000 (18:44 +0200)]
BUG7318: ETREE ODL learning leaf MACS on wrong tag
Change-Id: I1d65f6d70054eb4d48e6c0b1d158e9104b0c92d4
Signed-off-by: David Goldberg <gdavid@hpe.com>
Sridhar Gaddam [Wed, 11 Jan 2017 06:42:27 +0000 (12:12 +0530)]
Bug 7520: Avoid creating auto-tunnels for VLAN tenant networks
Currently auto-tunnel code of Netvirt is creating a Transport zone with
vteps even for tenant VLAN networks. This patch updates the code to create
the TZ only if the network type is VxLAN.
Change-Id: Iceeac4dc2dda99a41860d29f5d6f635f964b1a69
Signed-off-by: Sridhar Gaddam <sgaddam@redhat.com>
Vishal Thapar [Sun, 8 Jan 2017 10:34:19 +0000 (16:04 +0530)]
BUG 7488: Autobridge overwrites DpnId if bridge already exists
Change-Id: I49fcc00b116af84609e7f0654072a26f51fa6400
Signed-off-by: Vishal Thapar <vishal.thapar@ericsson.com>
VinothB [Thu, 22 Dec 2016 19:15:24 +0000 (00:45 +0530)]
Bug-7363: Fix for Flows are overlapped when we add custom SG along with ANY rule.
*Added reg match condition to avoid flow overlapping issue in other_protocols
Change-Id: Iea12c93a64b418a7e0028359ca54473bac76077b
Signed-off-by: VinothB <vinothb@hcl.com>
Sam Hague [Wed, 11 Jan 2017 18:44:56 +0000 (13:44 -0500)]
Use Developer Guide text in place of Documentation
Change-Id: I8ae9301af107f38d20ec0e0307d542eb5016335a
Signed-off-by: Sam Hague <shague@redhat.com>
Sam Hague [Tue, 10 Jan 2017 21:50:22 +0000 (16:50 -0500)]
Updates to NetVirt docs from some user feedback
Change-Id: I0f2ce58052d921813b1ae75530aa58eca4cd919d
Signed-off-by: Alon Kochba <alonko@hpe.com>
Signed-off-by: Sam Hague <shague@redhat.com>
Sam Hague [Tue, 10 Jan 2017 21:22:24 +0000 (16:22 -0500)]
Add basic initial docs for new layout
Change-Id: Ib6588368370f349e6695faa463cd3c847b2921f9
Signed-off-by: Sam Hague <shague@redhat.com>