Losing SNAT traffic when restarting schema-transformer

[Bruno Fernando]

I face that issue with contrail 3.2. Each time, that I restart the schema-transfomer, I'm losing few packets only f
or SNAT trafic.
What I have noticed so far is that the RI of the VN SI left is losing the logical-router RT at some point during the restart, and got it again after some time (also during the restart).

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/45596
Submitter: Bruno Fernando (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Reviewed: https://review.opencontrail.org/45596
Committed: http://github.com/Juniper/contrail-controller/commit/182f70500ca104b50b26bc27906939eba49e0109
Submitter: Zuul (<email address hidden>)
Branch: R3.2

commit 182f70500ca104b50b26bc27906939eba49e0109
Author: bfernando <email address hidden>
Date: Wed Aug 8 14:39:32 2018 +0200

schema-transfomer: don't update RT list when SI init isn't done

* When RoutingInstanceST are initialized, static routes are updated.
* This update trigger an update of route targets list which should not
be updated in the case where the RI is owned by an SNAT SI and that
latter has not been initialized yet in the reinit function.
* Prevent the schema to remove RT of SNAT's RI and so prevent to lose SNAT
traffic temporarily when restarting schema-transformer.

Closes-Bug: #1787371

Change-Id: I135eb89310d8cb0e34de245b363cbcb571a8eab7

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/48378
Submitter: Édouard Thuleau (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/48379
Submitter: Édouard Thuleau (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/48380
Submitter: Édouard Thuleau (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Reviewed: https://review.opencontrail.org/48380
Committed: http://github.com/Juniper/contrail-controller/commit/f19873501967a293d2a8719771fcd5153a17002c
Submitter: Zuul (<email address hidden>)
Branch: R4.1

commit f19873501967a293d2a8719771fcd5153a17002c
Author: bfernando <email address hidden>
Date: Wed Aug 8 14:39:32 2018 +0200

schema-transfomer: don't update RT list when SI init isn't done

* When RoutingInstanceST are initialized, static routes are updated.
* This update trigger an update of route targets list which should not
be updated in the case where the RI is owned by an SNAT SI and that
latter has not been initialized yet in the reinit function.
* Prevent the schema to remove RT of SNAT's RI and so prevent to lose SNAT
traffic temporarily when restarting schema-transformer.

Closes-Bug: #1787371

Change-Id: I135eb89310d8cb0e34de245b363cbcb571a8eab7
(cherry picked from commit 182f70500ca104b50b26bc27906939eba49e0109)

[OpenContrail Admin (ci-admin-f)]

Reviewed: https://review.opencontrail.org/48378
Committed: http://github.com/Juniper/contrail-controller/commit/d9cb298b7531dcb4723bfd8f062067811bfd2d0e
Submitter: Zuul v3 CI (<email address hidden>)
Branch: master

commit d9cb298b7531dcb4723bfd8f062067811bfd2d0e
Author: bfernando <email address hidden>
Date: Wed Aug 8 14:39:32 2018 +0200

schema-transfomer: don't update RT list when SI init isn't done

* When RoutingInstanceST are initialized, static routes are updated.
* This update trigger an update of route targets list which should not
be updated in the case where the RI is owned by an SNAT SI and that
latter has not been initialized yet in the reinit function.
* Prevent the schema to remove RT of SNAT's RI and so prevent to lose SNAT
traffic temporarily when restarting schema-transformer.

Closes-Bug: #1787371

Change-Id: I135eb89310d8cb0e34de245b363cbcb571a8eab7
(cherry picked from commit 182f70500ca104b50b26bc27906939eba49e0109)

[OpenContrail Admin (ci-admin-f)]

Reviewed: https://review.opencontrail.org/48379
Committed: http://github.com/Juniper/contrail-controller/commit/5fd37ce9c5876b15330a625cd7ab869c8da8efe2
Submitter: Zuul v3 CI (<email address hidden>)
Branch: R5.0

commit 5fd37ce9c5876b15330a625cd7ab869c8da8efe2
Author: bfernando <email address hidden>
Date: Wed Aug 8 14:39:32 2018 +0200

schema-transfomer: don't update RT list when SI init isn't done

* When RoutingInstanceST are initialized, static routes are updated.
* This update trigger an update of route targets list which should not
be updated in the case where the RI is owned by an SNAT SI and that
latter has not been initialized yet in the reinit function.
* Prevent the schema to remove RT of SNAT's RI and so prevent to lose SNAT
traffic temporarily when restarting schema-transformer.

Closes-Bug: #1787371

Change-Id: I135eb89310d8cb0e34de245b363cbcb571a8eab7
(cherry picked from commit 182f70500ca104b50b26bc27906939eba49e0109)

[OpenContrail Admin (ci-admin-f)]

Reviewed: https://review.opencontrail.org/49393
Committed: http://github.com/Juniper/contrail-controller/commit/77df3b58265b3fab414dfbc00e1ff39d19f0a99c
Submitter: Zuul v3 CI (<email address hidden>)
Branch: R6.0-WIP

commit 77df3b58265b3fab414dfbc00e1ff39d19f0a99c
Author: Shivayogi Ugaji <email address hidden>
Date: Mon Nov 5 22:07:18 2018 -0800

Apply commits from master onto R6.0-WIP

db_resync_done lock is used to indicate the amqp thread to wait for resync to
complete. In this case, when we call SchemaTransformer.destroy_instance()
due to Casandra connection failure, this lock remains locked blocking
destroy_instance. destroy_instance calls _vnc_subscribe_callback to drain the
amqp queue which waits infinitely for db_resync_done lock to be released.
This fix releases db_resync_done lock so that destroy_instance doesnt get
blocked.
Closes-Bug: #1801474

[DM] Hitless image upgrade implementation
Closes-Bug: #1799322

Provisioner for the devicemanager node.
usage:
from /opt/contrail/utils
python provision_devicemgr_node.py --host_name aio --host_ip 10.87.82.2
--oper add --admin_user admin --admin_password contrail123 --admin_tenant_name
admin --openstack_ip 10.87.82.2 --api_server_ip 10.87.82.2
Closes-Bug: #1805303

CFM: Changes for onboarding L3PNF
- Add new platform SRX240
- Add L3PNF subnet is schema
- Add new namespace, VN and IPAM for L3PNF during brownfield onboarding
Closes-Bug: 1800701

Add entrypoint to vrouter-agent service on Windows
Introduce entrypoint for agent similar in design to that from
microservice deployment. For now it will only start agent,
actual features will be added in following changes.
Partial-Bug: #1806677

Check build dependencies for tbb, SimpleAmqpClient and rabbitmq
Closes-Bug: #1806719

Make agent's entrypoint update agent's config on Windows
In future we will generate the whole config from scratch
as on Linux, but for now we only update the vhost's ifname.
It's the only field that can change upon restart.
Partial-Bug: #1806677

bgp-peer selection support for bgpaas
1. Listener BgpRouterConfig is added for BgpRouter and ControlNodeZone
2. BgpRouterConfig builds BgpRouterTree and ControlNodeZoneTree
from IFMapNode
3. BGPaaS gets BgpRouter for configured ControlNodeZone from
BgpRouterConfig and Updates bgp-peer-ip and bgp-peer-port in
the flow.
4. Step 3 is followed for xmpp based peer-selection also.
5. BGPaaS sandesh is updated with primary_control_node_zone,
secondary_control_node_zone, bgp_peer_ip and bgp_peer_port
Partial-bug: #1775872

[DM] Inside-outside workflow - lag/mH
1. Change the exisiting business logic to adhere to the new data model for lag/mH workflow
2. Multi-vlan support
Partial-Bug: #1799329

Rework nodemgr before fixing ntp issue
- move windows/linux code to separate classes instead of same condition through the code
- simplify main.py
- remove copy duplication
Closes-Bug: 1800704

[fabric] Added playbook retry support to job manager
1) When playbook return retry_devices in the output, job manager will retry the playbooks against those devices
2) remove obsolete playbooks from 5.0
3) remove obsolete ansible roles from 5.0
4) added a warning log on missing loopback interface when ...