Change-Id: I5f22cdb4d021eb6f16ccbcd900fba4cdf63fe4a9

Change-Id: If61f183236d16a1c10181284e63d42f5d2eee9b0

Change-Id: Id519f13a37d2252255b7c2baa0fe103eeefe8ed4
Closes-Bug: 1685954
Closes-Bug: 1685040

https://tools.ietf.org/html/draft-ietf-idr-bgp-gr-notification-10

Attempt GR Helper mode for all received notifications (and hold-timer expiry)
except HardReset and PeerDeconfigured.

Always send Notifications. NotificationFlag support has been in JUNOS already
for a while. Older versions of control-node already skips most of the
notifications received (and triggers GR helper mode)

Also fix a couple of nits found by cpplint

Closes-Bug: 1680047
Change-Id: Ic0462e9483dda8ee2f6c1be503bf1c120c3d6786
(cherry picked from commit c45ed8b)

receives body without 'resource' key. Simulate one and ensure any
port association is cleared in such a case.

Change-Id: I7ab14cb45b4896ddd4039f1343178d1984a4ca13
Closes-Bug: 1685314
(cherry picked from commit 4a2080d)

If the IP address is different, then it implies mis-configuration of
the xmpp agents. In such cases, drop new connection request and
retain existing connection intact

Since xmpp tasks run concurrently (across different IP addresses),
use a XmppConnectionManager scoped mutex to solve concurrency issues
among such parallel xmpp connection requests processing

Also increase severity of a couple of related log messages

Re-enable and fix few UTs that test this very exact scenario

Closes-Bug: #1687096

Change-Id: Ic7252060abfc4b34fa184ff163afc217a88bb40c

If all policies are removed from a network, we were not setting
the acl counts back to 0. We should initialize it to 0 before
evaluating policies.

Change-Id: Ie1575095ca117346b3deefa7fe7d22e0ed59c492
Closes-Bug: 1685458
(cherry picked from commit cb6f312)

Config: Address pairs should not be allowed when port security
        is not enabled

Change-Id: I211bd7e89d82633bb6701ddb55d1527633ea246d
Closes-Bug: 1685030

to route-target for any configured route-target values.

Change-Id: Ida9bb18c40bf5576fc11c1dcb536b6a3db806f5a
Closes-Bug: 1687824

When a UVE object gets deleted, any alarms associated with that UVE object
also gets deleted. However, the corresponding UVE key is not removed from the
alarm table in contrail-alarm-gen, resulting in duplicate alarm delete
messages being sent continuously.

Change-Id: I598f58687179d82c11f531b337508385098c9036
Closes-Bug: #1615782

Change-Id: I3d60d02f230ccc188ebbda58701423d5120024e4
Closes-Bug: 1685734

Send VN UUID, vlan tags and parent interface name in VMI UVE
Closes-Bug: #1673490

Change-Id: I47597fb83278d4f322e5d7257576381e86a77e2f

arriving in TCP so taking off the support. The reason being, unlike
UDP, TCP sockets read multiple syslog messages in one read until
the buffer gets full. This can toss our parsing code.
Closes-Bug:#1687475

Change-Id: I87c72e2154e5eecd19657a1de92c6c28b77fcd35

Problem:
On SI compute routes are imported from CN with ECMP NH.
These routes were not notifying change in load balance params if changed.

Solution:
Notify on change in ecmp hash fields.

Note:
Any change in hash fields gets applied to new flows. Existing flows continue to
use same index as computed before.

Closes-bug: #1643842

Conflicts:
	src/vnsw/agent/controller/controller_route_path.cc
Change-Id: I0216e12ba90b152c58deaaf67b477876eddfc9e8

Added description field for neutron objects.

Change-Id: I79969fb1b3b026bb0480d94900de72a9acf8372a
Closes-Bug: 1686476
Closes-Bug: 1685940

Updating default SG should not be allowed and should raise
a 'SecurityGroupCannotUpdateDefault' neutron exception.

We are only checking for this in neutron_plugin_db.py and
will still allow the update to default SG from vnc_api.

Closes-Bug: 1686496

Change-Id: I7eab47b0047dd2e35ec270572d75d38eac3d90bc

o Send GR capability even when configured restart-time is 0
o Enable GR helper even if received restart-time is 0

Fix a few issues in LLGR

o Process received notification code-subcode correctly and trigger
  GR/LLGR Helper mode accordingly
o Add UTs to cover most of the scenarios described above

Change-Id: Ibcef9c059408a61e8513b6a9f6e1d015128e6515
Closes-Bug: 1688133

When accessing control-node introspect,
bgp standard communities does not show up
that are configured as part of static routes.
It is seen that extended communities associated
with static routes override the bgp standard communities
in introspect.
Modify method FillRoutePathExtCommunityInfo
in /opt/stack/contrail/controller/src/bgp/bgp_route.cc
to avoid overwriting bgp standard communities.

Change-Id: I13f2d2b338a5464afd56ebc558036fea507b2394
Closes-Bug:1655088

* When both filters (network-id and device-id) are used, the
  vnc_openstack has to compare the device-id and the network-id. Both
  have to be present in the vmi_obj refs. The vnc_lib with both refs returns
  a result with an OR logical operation, therefore with the current comparison
  the vnc_openstack returns a result with an AND on both filters.
* Add a new conditional entry that manages the case if there is only the
  network-id provided in the filters.

Change-Id: I3054345a84caeb892e85a350533e4e600f353665
Closes-Bug: #1685764
(cherry picked from commit 8ff6c71)

Following changes are implemented:

- Use a vector instead of an array for BgpPeer buffer
- Make the vector capacity configurable via an env variable
- Add unit tests to exercise the new code

Change-Id: I6490dad8797b23dfb2d65a4681c1e1f663e0dd85
Closes-Bug: 1687442

Made keystone auth params optional so that it works for keystone and no-auth cases.

Change-Id: I83e255d057621d217d305c545285032a85bf2f59
Closes-Bug: #1677070
(cherry picked from commit cffa0bc)

These lists do not include BGPaaS peers.

Change-Id: Ie1a359bb5c1b84e6cb93cbbb6e7d0c05b44fd694
Closes-Bug: 1638065

Change-Id: I490a0919b8ffa8b07c91ee324a10569f7901109c
Closes-Bug: 1683548

Change-Id: I31f4e35d5588a746bd896ee06314f4680454eaf9
Closes-Bug: #1683672
(cherry picked from commit 9c182b4e57f78e8be48a7c890fa5ef019db2d223)

Added a check to see if router's and (port/subnet)'s tenant ids are
same at the time of adding an interface to a router.

Closes-Bug: 1686489

Change-Id: Ibc2fe74c6ce56e467aec87698077d145751f64cc

…k receives body without 'resource' key. Simulate one and ensure any port association is cleared in such a case." into R3.2.3.x

…e" into R3.2.3.x

…ref to route-target for any configured route-target values." into R3.2.3.x

…arriving in TCP so taking off the support. The reason being, unlike UDP, TCP sockets read multiple syslog messages in one read until the buffer gets full. This can toss our parsing code. Closes-Bug:#1687475" into R3.2.3.x

… R3.2.3.x

…UVE" into R3.2.3.x

…2.3.x

uses it. In 4.0 these libraries were added as part of securing
introspect changes, so no build failures seen. But for 3.2 they
have to be added.
Closes-Bug:#1695543

Change-Id: I6f2644c3fb5d4c28eddc61d05966efd3d13d15cd

…fka uses it. In 4.0 these libraries were added as part of securing introspect changes, so no build failures seen. But for 3.2 they have to be added. Closes-Bug:#1695543" into R3.2.3.x

contrail-svc-monitor uses the uve status if the query
is succeed to the contrail-analytics. Otherwise it is
better to continue with last known state of agents and vrouters

Change-Id: I3b4c896a8834f5fef71ff4262dc09dc62cee36bc
Closes-bug: #1696593

ConnectionPool initialization object.
Which will retry (max_retries) time with the sleep
of "_BASE_BACKOFF * (2 ** self._retry_count)" seconds

Where _BASE_BACKOFF = 0.01 and self._retry_count can
reach up to 30(max_retries), during an operation on a
connection fails due to a TimedOutException or
UnavailableException

changing the max_reties to 15 so that the max sleep
time will be approximately 5 minutes before the config
services are restarted due to DatabaseUnavailableError

Change-Id: Ib96c9e3f9745bcb06fea621e0d7fb52c8ff87f69
Closes-Bug: 1697252
(cherry picked from commit e31027f)

class handle method.

Change-Id: I6b4ba7c20e9b13afa62704fb879e69550c68ad17
Closes-Bug: 1697548
(cherry picked from commit 37ba4e9)

Change-Id: Ifcf3d3c81dccfbda808c2dc747290a05963d4de9

Authorize administrative users to set router interface even the router
does not belongs to the user's project.

Change-Id: Ie9f44b60a9018f67bd62a49b00aef22a6ec77ac7
Partial-Bug: #1686489
(cherry picked from commit 3bea664)
(cherry picked from commit 4b37571)

Problem:
Fat flow is enabled on the vm-interface and not on vhost0 interface.
So, packets from vhost0 result in non-fat flow and packets from
vm-interface result in fat-flow.

Health Check assumes that flows are always initiated from vhost0
interface. So, flows created due to packets from vm-interface do not
have NAT translation based on health-check rules. As a result, packets
from vm-interface are going to host-os

Solution:

1. Health-Check status is ACTIVE
When fat-flow is enabled, the fat-flow created due to packet from
vm-interface will also apply health-check rules and create appropriated
NAT-ed reverse flow.

2. Health-Check status is INACTIVE
When health-check status is INACTIVE, the routes for IP address on
vm-interface are also withdrawn. So, flows cannot be setup for packets
from vm-interface.
As a solution, we disable fat-flow on the vm-interface when health-check
status is INACTIVE. As a result, packets from vm-interface will match
flows created due to request and health-check session will succeed.

Conflicts:
	src/vnsw/agent/pkt/test/SConscript

Change-Id: I7766cbd9bbc25e1f740148c35ed7e115a4fd7d95
Closes-Bug: #1676721

If contrail-dns doesnt generate global options and named reads the
config file in the time between contrail-dns generation and script
update, it woulnt listen on port 8094.

Change-Id: Iabb6203bbde1d900b7f20bbe2330373e001ff300
closes-bug: #1694927
(cherry picked from commit 7ee4969)

Agent oper tables point the GW and DNS routes to pkt0 using NH with policy
enabled. However, this is not being updated in vrouter due to an additional
check in ksync. Removing the same so that traffic sent to pkt0 will always
have flows setup.

Change-Id: Icf3fb0be8bda3f8d482db3ecc7356a20d43a18cd
closes-bug: #1636574

… disabled" into R3.2.3.x

Change-Id: Ia60c141ccdaabef814cc97ccb28be3db7bb6e969
closes-bug: #1699983

when request comes for the update where user has entered uppercase
gw_ip or dns_server_address, string compare would fail, causing
network_update to fail. Modify the validate check to compare lower case
strings.
Closes-Bug: #1667155

Change-Id: Iae6d0eb2621fc4d960beb0c643f3877a0452dfd1

Change-Id: Ief5a66f266bd6accb7d99e0d06633bf98445e4f7
Closes-Bug: 1697530
(cherry picked from commit cc5e1b9)

_() was wrongly used for contrail exception messages in neutron_plugin_db.
It is not imported nor the translations are available, so it was raising
different exception without meaningful message.

Change-Id: I07a20c66980deda0b053beaa691d9ea8e4c75d9b
Partial-Bug: #1700622

Fix ZooDelete in zookeeper client to retry recoverable errors

Closes-Bug #1692453

Change-Id: If9a9d10d8675acd181aeb2f8436c1aa944cbb13f
(cherry picked from commit 97b9d19)

…2.3.x

Change-Id: Ia9abb2540ebbf32d3e9e142a9ebfd7f54269628a

Change-Id: I0453a360b2f2708b1e48a93e35ddc4d241bf89db
Closes-Bug: #1708239

This fix adds a new API cass_value_is_null to our CQL library so
that we can check if a value is null or not before trying to decode
the datatype and readh the column value.

Re-enable opserver:test

Closes-Bug: #1699934

Conflicts:
	src/database/cassandra/cql/cql_if.cc
	src/database/cassandra/cql/cql_lib_if.h

(cherry picked from commit 5b2442c)
(cherry picked from commit 1f6331a)

Change-Id: I884f4c275c7cb92b19884d4cc4d45f9cd3a2c8d4

In case of short lived TCP session using same 5 tuple we might
end up accesssing a reverse flow entry which might be already deleted
and present in free list. Below set of events lead to this issue

1. F1 and R1 flow are added to flow table
2. R1 is written to vrouter
3. F1 is written to vrouter
4. R1 flow add response is received, triggering update of
   F1(not needed now as reverse flow index is not written to
   kernel?)
5. In the meantime flow is evicted in vrouter, hence flow
   update for F1 would result in error from vrouter resulting in short flow
6. Since F1 is shortflow Flow delete gets enqueued
7. Since R1 is evict marked, flow evict gets enqueued
8. Both event F1 and R1 delete and evict event can run in
   parallel,and hence reverse flow pointer obtained before FLOW
   lock could be invalid, hence read back the same

Was able to simulate the same in UT, UT is disabled as it need some
instrumentation to run it to hit the issue

Change-Id: I368f465f9d446b43dfe0e4baa547d8c8cbfd6840
Closes-bug: #1714371

the new librdkafka requires collector to be
linked against these new libraries.
Closes-Bug:#1711833

Change-Id: Ie4f6e06372be6292a9834795409a09d7aea02990

Avoid using old cached subnet when re-creating subnet with same
CIDR but different gateway setting.

Change-Id: I51b25c946b75a72fb0139e9250f319389588b380
Closes-Bug: #1722837

This change allows filter port list by mac_address field.

Closes-Bug: 1718447

Change-Id: If2425f8553701440eb4fedd0b391268752b237ed
(cherry picked from commit e437fde)

Updating vnic type is legal procedure, and is used by Ironic. This patch
allows to update vnic type when port is not bound (doesn't have
references to VM or Vrouter)

Closes-Bug: 1719268

Change-Id: Ic7b6a7232316c698b2f8521a9b3d46fb755d4e3d
(cherry picked from commit 3ef7d84)

…d in alarmgen.

Closes-Bug: #1697745

Change-Id: I1861c2415f05f0cd71d52172f422afff15abc821

Conflicts:
	src/opserver/alarmgen.py

Change-Id: I1861c2415f05f0cd71d52172f422afff15abc821

Conflicts:
	src/opserver/alarmgen.py
Change-Id: I1861c2415f05f0cd71d52172f422afff15abc821

The backref entries are updated in DB and ksync tasks, which are mutually
exclusive. The flow event task was accessing the same and it doesnt actually
require. Avoiding backref reevaluation.

Change-Id: I8cb9b92196626ea219f7f9e1055e2579c62c7c86
closes-bug: #1726709
(cherry picked from commit ef13c6b)

Agent Profile task is updating maps in the profile data (like
profile_stats_table_) while sandesh is fetching the same and iterating them,
resulting in crash.

Change-Id: I0503035340fd505610e8114f5341c5289ecdda12
closes-bug: #1719840
(cherry picked from commit 540592c)

Flow index passed in ECMP resolve trap could be -1 resulting in no
flow entry being present in agent. In such case agent was accessing a
invalid flow pointer, initializing the same.

Change-Id: I559b6be8f75bae87442517e6edab00868742e397
Closes-bug: #1717507
(cherry picked from commit da656c5)
(cherry picked from commit 45e8e93)

A flow is marked as reverse flow and its forward flow is NULL. Recompute flow
is accessing this forward flow causing the crash. Updating to ignore this case.

Change-Id: I08ac5beb3cc6b1879d8679d7a582ecf1eedb2286
closes-bug: #1719857
(cherry picked from commit 41779f4)

Flow addition and deletion to flow table happens in same task context,
addition happens in flow_event work queue and deletion from flow tree
and freeing of flow entry happens in tokenless_queue. For every flow
add message processed we would have 2 flow entry created, hence free
queue should process twice the no. of entries so as to relinquish
free entries inline with flow addition rate.

Change-Id: Ic99ced407594dec50b998d80e8e52b08876783f3
Closes-bug: #1728316
(cherry picked from commit 2030642)

When we have dual interfaces, v4 and v6 service health check addresses are
created and sent to agent. Agent oper data has only one service health check
address, which is a problem. Updating to use only the v4 address. Need to add
support for v6 health check separately.

Change-Id: Ia5c1973c500d1e6d5d89ca83c3a48b8d94c0e579
closes-bug: #1692283
(cherry picked from commit 93d5d4e)

If mac address of a port is chnaged via port-update API,
it is ignored. This fix handles the updating of the mac
address. Also, it ensures that the same mac address
does not exixt on any other port in the same virtual
network.

Change-Id: Iec9290e0bc4485f2d59c38b223ddc458679aa1f4
Closes-Bug: 1682672
Partial-Bug: 1683547
(cherry picked from commit 3905546)

… R3.2.3.x

…R3.2.3.x

….2.3.x

…valuated in alarmgen. Closes-Bug: #1697745" into R3.2.3.x

Change-Id: I0c7821695cccffb60496a55b6e7af4a27ebfe5e2
Closes-Bug: #1747427

Change-Id: I9fa3f3f494ec2769cc03ec88987aa68903643999
Closes-Bug: #1735409
(cherry picked from commit ccdaf51)
(cherry picked from commit a6dfe12)