openstack
/
octavia
Code Issues Proposed changes

Remove v1 API and associated code 

Includes some updates to docs and configs and related files to remove
references to neutron-lbaas. Also remove handlers.

Change-Id: I3082962841d3b645f3cbd1a6b41fc7fb28dcf7e6
changes/07/656707/4
Adam Harwell 2019-05-01 14:12:22 -07:00
parent 1a87298ac7
commit 29d4340e9f
110 changed files with 85 additions and 16736 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
api-ref/source/examples/versions-get-resp.json
View File

@ -1,13 +1,5 @@
{
"versions": [{
"status": "DEPRECATED",
"updated": "2014-12-11T00:00:00Z",
"id": "v1",
"links": [{
"href": "http://10.21.21.53/load-balancer/v1",
"rel": "self"
}]
}, {
"status": "SUPPORTED",
"updated": "2016-12-11T00:00:00Z",
"id": "v2.0",

5
api-ref/source/index.rst
View File

@ -15,15 +15,10 @@ Supported API version
None
Deprecated API version
:doc:`v1/octaviaapi`
.. toctree::
:hidden:
v2/index
v1/octaviaapi
Octavia API minor releases are additive to the API major revision and share
the same URL path. Minor revision changes to the API are called out in the API

2280
api-ref/source/v1/octaviaapi.rst
View File

File diff suppressed because it is too large Load Diff

9
devstack/plugin.sh
View File

@ -270,15 +270,6 @@ function octavia_configure {
iniset $OCTAVIA_CONF oslo_messaging rpc_thread_pool_size 2
iniset $OCTAVIA_CONF oslo_messaging topic octavia_prov
# TODO(nmagnezi): Remove this when neutron-lbaas gets deprecated
# Setting neutron request_poll_timeout
iniset $NEUTRON_CONF octavia request_poll_timeout 3000
if [[ "$WSGI_MODE" == "uwsgi" ]]; then
iniadd $NEUTRON_CONF octavia base_url "$OCTAVIA_PROTOCOL://$SERVICE_HOST/$OCTAVIA_SERVICE_TYPE"
else
iniadd $NEUTRON_CONF octavia base_url "$OCTAVIA_PROTOCOL://$SERVICE_HOST:$OCTAVIA_PORT/"
fi
# Uncomment other default options
iniuncomment $OCTAVIA_CONF haproxy_amphora base_path
iniuncomment $OCTAVIA_CONF haproxy_amphora base_cert_dir

4
devstack/settings
View File

@ -90,11 +90,7 @@ GITREPO["octavia-lib"]=${OCTAVIA_LIB_REPO:-${GIT_BASE}/openstack/octavia-lib.git
GITBRANCH["octavia-lib"]=${OCTAVIA_LIB_BRANCH:-master}
GITDIR["octavia-lib"]=$DEST/octavia-lib
NEUTRON_LBAAS_DIR=$DEST/neutron-lbaas
NEUTRON_LBAAS_CONF=$NEUTRON_CONF_DIR/neutron_lbaas.conf
OCTAVIA_SERVICE_PROVIDER=${OCTAVIA_SERVICE_PROVIDER:-"LOADBALANCERV2:Octavia:neutron_lbaas.drivers.octavia.driver.OctaviaDriver:default"}
NEUTRON_ANY=${NEUTRON_ANY:-"q-svc neutron-api"}
LBAAS_V2=${LBAAS_V2:-"neutron-lbaasv2"}
# HA-deployment related settings
OCTAVIA_USE_PREGENERATED_SSH_KEY=${OCTAVIA_USE_PREGENERATED_SSH_KEY:-"False"}

2
doc/source/conf.py
View File

@ -1,6 +1,6 @@
# -*- coding: utf-8 -*-
#
# Tempest documentation build configuration file, created by
# Octavia documentation build configuration file, created by
# sphinx-quickstart on Tue May 21 17:43:32 2013.
#
# This file is execfile()d with the current directory set to its containing

56
doc/source/contributor/guides/dev-quick-start.rst
View File

@ -72,8 +72,7 @@ Deployment
2. Copy ``devstack/contrib/new-octavia-devstack.sh`` from this source
repository onto that host.
3. Run new-octavia-devstack.sh as root.
4. Deploy loadbalancers, listeners, etc. as you would with any Neutron LBaaS v2
enabled cloud.
4. Deploy loadbalancers, listeners, etc.
Running Octavia in production
@ -125,7 +124,7 @@ For the purposes of this guide, we will therefore assume the following core
components have already been set up for your production OpenStack environment:
* Nova
* Neutron (with Neutron LBaaS v2)
* Neutron
* Glance
* Barbican (if TLS offloading functionality is enabled)
* Keystone
@ -138,11 +137,8 @@ Production Deployment Walkthrough
Create Octavia User
___________________
By default Octavia will use the 'neutron' user for keystone authentication, and
the admin user for interactions with all other services. However, it doesn't
actually share neutron's database or otherwise access Neutron outside of
Neutron's API, so a dedicated 'octavia' keystone user should generally be
created for Octavia to use.
By default Octavia will use the 'octavia' user for keystone authentication, and
the admin user for interactions with all other services.
You must:
@ -225,14 +221,8 @@ Running multiple instances of the individual Octavia controller components on
separate physical hosts is recommended in order to provide scalability and
availability of the controller software.
One important security note: In 0.9 of Octavia, the Octavia API is designed to
be consumed only by the Neutron-LBaaS v2 Octavia driver. As such, there is
presently no authentication required to use the Octavia API, and therefore the
Octavia API should only be accessible on trusted network segments
(specifically, the segment that runs the neutron-services daemons.)
The Octavia controller presently consists of several components which may be
split across several physical machines. For the 0.9 release of Octavia, the
split across several physical machines. For the 4.0 release of Octavia, the
important (and potentially separable) components are the controller worker,
housekeeper, health manager and API controller. Please see the component
diagrams elsewhere in this repository's documentation for detailed descriptions
@ -253,7 +243,7 @@ components need access to outside resources:
| housekeeper | Yes | Yes | No |
+-------------------+------------+----------+----------------+
In addition to talking to each other via OSLO messaging, various controller
In addition to talking to each other via Oslo messaging, various controller
components must also communicate with other OpenStack components, like nova,
neutron, barbican, etc. via their APIs.
@ -438,46 +428,22 @@ You must:
* Make sure each Octavia controller component is started appropriately.
Configuring Neutron LBaaS
_________________________
This is fairly straightforward. Neutron LBaaS needs to be directed to use the
Octavia service provider. There should be a line like the following in
``/etc/neutron/neutron_lbaas.conf`` file's ``[service providers]`` section:
::
service_provider = LOADBALANCERV2:Octavia:neutron_lbaas.drivers.octavia.driver.OctaviaDriver:default
In addition to the above you must add the octavia API ``base_url`` to the
``[octavia]`` section of ``/etc/neutron/neutron.conf``. For example:
::
[octavia]
base_url=http://127.0.0.1:9876
You must:
* Update ``/etc/neutron/neutron_lbaas.conf`` as described above.
* Add the octavia API URL to ``/etc/neutron/neutron.conf``.
Install Neutron-LBaaS v2 extension in Horizon
Install Octavia extension in Horizon
_____________________________________________
This isn't strictly necessary for all cloud installations, however, if yours
makes use of the Horizon GUI interface for tenants, it is probably also a good
idea to make sure that it is configured with the Neutron-LBaaS v2 extension.
idea to make sure that it is configured with the Octavia extension.
You may:
* Install the neutron-lbaasv2 GUI extension in Horizon
* Install the octavia GUI extension in Horizon
Test deployment
_______________
If all of the above instructions have been followed, it should now be possible
to deploy load balancing services using the python neutronclient CLI,
communicating with the neutron-lbaas v2 API.
to deploy load balancing services using the OpenStack CLI,
communicating with the Octavia v2 API.
Example:

672
doc/source/user/guides/basic-cookbook-neutron.rst
View File

@ -1,672 +0,0 @@
..
Copyright (c) 2016 IBM
Licensed under the Apache License, Version 2.0 (the "License"); you may
not use this file except in compliance with the License. You may obtain
a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
License for the specific language governing permissions and limitations
under the License.
===============================================================
Basic Load Balancing Cookbook Using Neutron Client (deprecated)
===============================================================
.. warning:: The neutron client used in this document is deprecated. We
strongly encourage you to use the OpenStack Client and Octavia
OpenStack Client plugin instead. This document is being maintained
for deployments still using neutron-lbaas and the neutron client.
Introduction
============
This document contains several examples of using basic load balancing services
as a tenant or "regular" cloud user.
For the purposes of this guide we assume that the neutron and barbican
command-line interfaces are going to be used to configure all features of
Neutron LBaaS with an Octavia back-end. In order to keep these examples short,
we also assume that tasks not directly associated with deploying load balancing
services have already been accomplished. This might include such things as
deploying and configuring web servers, setting up Neutron networks, obtaining
TLS certificates from a trusted provider, and so on. A description of the
starting conditions is given in each example below.
Please also note that this guide assumes you are familiar with the specific
load balancer terminology defined in the :doc:`../../reference/glossary`. For a
description of load balancing itself and the Octavia project, please see:
:doc:`../../reference/introduction`.
Examples
========
Deploy a basic HTTP load balancer
---------------------------------
While this is technically the simplest complete load balancing solution that
can be deployed, we recommend deploying HTTP load balancers with a health
monitor to ensure back-end member availability. See
:ref:`basic-lb-with-hm-neutron` below.
**Scenario description**:
* Back-end servers 192.0.2.10 and 192.0.2.11 on subnet *private-subnet* have
been configured with an HTTP application on TCP port 80.
* Subnet *public-subnet* is a shared external subnet created by the cloud
operator which is reachable from the internet.
* We want to configure a basic load balancer that is accessible from the
internet, which distributes web requests to the back-end servers.
**Solution**:
1. Create load balancer *lb1* on subnet *public-subnet*.
2. Create listener *listener1*.
3. Create pool *pool1* as *listener1*'s default pool.
4. Add members 192.0.2.10 and 192.0.2.11 on *private-subnet* to *pool1*.
**CLI commands**:
::
neutron lbaas-loadbalancer-create --name lb1 public-subnet
# Re-run the following until lb1 shows ACTIVE and ONLINE statuses:
neutron lbaas-loadbalancer-show lb1
neutron lbaas-listener-create --name listener1 --loadbalancer lb1 --protocol HTTP --protocol-port 80
neutron lbaas-pool-create --name pool1 --lb-algorithm ROUND_ROBIN --listener listener1 --protocol HTTP
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.10 --protocol-port 80 pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.11 --protocol-port 80 pool1
.. _basic-lb-with-hm-neutron:
Deploy a basic HTTP load balancer with a health monitor
-------------------------------------------------------
This is the simplest recommended load balancing solution for HTTP applications.
This solution is appropriate for operators with provider networks that are not
compatible with Neutron floating-ip functionality (such as IPv6 networks).
However, if you need to retain control of the external IP through which a load
balancer is accessible, even if the load balancer needs to be destroyed or
recreated, it may be more appropriate to deploy your basic load balancer using
a floating IP. See :ref:`basic-lb-with-hm-and-fip-neutron` below.
**Scenario description**:
* Back-end servers 192.0.2.10 and 192.0.2.11 on subnet *private-subnet* have
been configured with an HTTP application on TCP port 80.
* These back-end servers have been configured with a health check at the URL
path "/healthcheck". See :ref:`http-heath-monitors-neutron` below.
* Subnet *public-subnet* is a shared external subnet created by the cloud
operator which is reachable from the internet.
* We want to configure a basic load balancer that is accessible from the
internet, which distributes web requests to the back-end servers, and which
checks the "/healthcheck" path to ensure back-end member health.
**Solution**:
1. Create load balancer *lb1* on subnet *public-subnet*.
2. Create listener *listener1*.
3. Create pool *pool1* as *listener1*'s default pool.
4. Create a health monitor on *pool1* which tests the "/healthcheck" path.
5. Add members 192.0.2.10 and 192.0.2.11 on *private-subnet* to *pool1*.
**CLI commands**:
::
neutron lbaas-loadbalancer-create --name lb1 public-subnet
# Re-run the following until lb1 shows ACTIVE and ONLINE statuses:
neutron lbaas-loadbalancer-show lb1
neutron lbaas-listener-create --name listener1 --loadbalancer lb1 --protocol HTTP --protocol-port 80
neutron lbaas-pool-create --name pool1 --lb-algorithm ROUND_ROBIN --listener listener1 --protocol HTTP
neutron lbaas-healthmonitor-create --delay 5 --max-retries 4 --timeout 10 --type HTTP --url_path /healthcheck --pool pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.10 --protocol-port 80 pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.11 --protocol-port 80 pool1
.. _basic-lb-with-hm-and-fip-neutron:
Deploy a basic HTTP load balancer using a floating IP
-----------------------------------------------------
It can be beneficial to use a floating IP when setting up a load balancer's VIP
in order to ensure you retain control of the IP that gets assigned as the
floating IP in case the load balancer needs to be destroyed, moved, or
recreated.
Note that this is not possible to do with IPv6 load balancers as floating IPs
do not work with IPv6. Further, there is currently a bug in Neutron Distributed
Virtual Routing (DVR) which prevents floating IPs from working correctly when
DVR is in use. See: https://bugs.launchpad.net/neutron/+bug/1583694
**Scenario description**:
* Back-end servers 192.0.2.10 and 192.0.2.11 on subnet *private-subnet* have
been configured with an HTTP application on TCP port 80.
* These back-end servers have been configured with a health check at the URL
path "/healthcheck". See :ref:`http-heath-monitors-neutron` below.
* Neutron network *public* is a shared external network created by the cloud
operator which is reachable from the internet.
* We want to configure a basic load balancer that is accessible from the
internet, which distributes web requests to the back-end servers, and which
checks the "/healthcheck" path to ensure back-end member health. Further, we
want to do this using a floating IP.
**Solution**:
1. Create load balancer *lb1* on subnet *private-subnet*.
2. Create listener *listener1*.
3. Create pool *pool1* as *listener1*'s default pool.
4. Create a health monitor on *pool1* which tests the "/healthcheck" path.
5. Add members 192.0.2.10 and 192.0.2.11 on *private-subnet* to *pool1*.
6. Create a floating IP address on *public-subnet*.
7. Associate this floating IP with the *lb1*'s VIP port.
**CLI commands**:
::
neutron lbaas-loadbalancer-create --name lb1 private-subnet
# Re-run the following until lb1 shows ACTIVE and ONLINE statuses:
neutron lbaas-loadbalancer-show lb1
neutron lbaas-listener-create --name listener1 --loadbalancer lb1 --protocol HTTP --protocol-port 80
neutron lbaas-pool-create --name pool1 --lb-algorithm ROUND_ROBIN --listener listener1 --protocol HTTP
neutron lbaas-healthmonitor-create --delay 5 --max-retries 4 --timeout 10 --type HTTP --url_path /healthcheck --pool pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.10 --protocol-port 80 pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.11 --protocol-port 80 pool1
neutron floatingip-create public
# The following IDs should be visible in the output of previous commands
neutron floatingip-associate <floating_ip_id> <load_balancer_vip_port_id>
Deploy a basic HTTP load balancer with session persistence
----------------------------------------------------------
**Scenario description**:
* Back-end servers 192.0.2.10 and 192.0.2.11 on subnet *private-subnet* have
been configured with an HTTP application on TCP port 80.
* The application is written such that web clients should always be directed to
the same back-end server throughout their web session, based on an
application cookie inserted by the web application named 'PHPSESSIONID'.
* These back-end servers have been configured with a health check at the URL
path "/healthcheck". See :ref:`http-heath-monitors-neutron` below.
* Subnet *public-subnet* is a shared external subnet created by the cloud
operator which is reachable from the internet.
* We want to configure a basic load balancer that is accessible from the
internet, which distributes web requests to the back-end servers, persists
sessions using the PHPSESSIONID as a key, and which checks the "/healthcheck"
path to ensure back-end member health.
**Solution**:
1. Create load balancer *lb1* on subnet *public-subnet*.
2. Create listener *listener1*.
3. Create pool *pool1* as *listener1*'s default pool which defines session
persistence on the 'PHPSESSIONID' cookie.
4. Create a health monitor on *pool1* which tests the "/healthcheck" path.
5. Add members 192.0.2.10 and 192.0.2.11 on *private-subnet* to *pool1*.
**CLI commands**:
::
neutron lbaas-loadbalancer-create --name lb1 public-subnet
# Re-run the following until lb1 shows ACTIVE and ONLINE statuses:
neutron lbaas-loadbalancer-show lb1
neutron lbaas-listener-create --name listener1 --loadbalancer lb1 --protocol HTTP --protocol-port 80
neutron lbaas-pool-create --name pool1 --lb-algorithm ROUND_ROBIN --listener listener1 --protocol HTTP --session-persistence type=APP_COOKIE,cookie_name=PHPSESSIONID
neutron lbaas-healthmonitor-create --delay 5 --max-retries 4 --timeout 10 --type HTTP --url_path /healthcheck --pool pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.10 --protocol-port 80 pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.11 --protocol-port 80 pool1
Deploy a TCP load balancer
--------------------------
This is generally suitable when load balancing a non-HTTP TCP-based service.
**Scenario description**:
* Back-end servers 192.0.2.10 and 192.0.2.11 on subnet *private-subnet* have
been configured with an custom application on TCP port 23456
* Subnet *public-subnet* is a shared external subnet created by the cloud
operator which is reachable from the internet.
* We want to configure a basic load balancer that is accessible from the
internet, which distributes requests to the back-end servers.
* We want to employ a TCP health check to ensure that the back-end servers are
available.
**Solution**:
1. Create load balancer *lb1* on subnet *public-subnet*.
2. Create listener *listener1*.
3. Create pool *pool1* as *listener1*'s default pool.
4. Create a health monitor on *pool1* which probes *pool1*'s members' TCP
service port.
5. Add members 192.0.2.10 and 192.0.2.11 on *private-subnet* to *pool1*.
**CLI commands**:
::
neutron lbaas-loadbalancer-create --name lb1 public-subnet
# Re-run the following until lb1 shows ACTIVE and ONLINE statuses:
neutron lbaas-loadbalancer-show lb1
neutron lbaas-listener-create --name listener1 --loadbalancer lb1 --protocol TCP --protocol-port 23456
neutron lbaas-pool-create --name pool1 --lb-algorithm ROUND_ROBIN --listener listener1 --protocol TCP
neutron lbaas-healthmonitor-create --delay 5 --max-retries 4 --timeout 10 --type TCP --pool pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.10 --protocol-port 80 pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.11 --protocol-port 80 pool1
Deploy a non-terminated HTTPS load balancer
-------------------------------------------
A non-terminated HTTPS load balancer acts effectively like a generic TCP load
balancer: The load balancer will forward the raw TCP traffic from the web
client to the back-end servers without decrypting it. This means that the
back-end servers themselves must be configured to terminate the HTTPS
connection with the web clients, and in turn, the load balancer cannot insert
headers into the HTTP session indicating the client IP address. (That is, to
the back-end server, all web requests will appear to originate from the load
balancer.) Also, advanced load balancer features (like Layer 7 functionality)
cannot be used with non-terminated HTTPS.
**Scenario description**:
* Back-end servers 192.0.2.10 and 192.0.2.11 on subnet *private-subnet* have
been configured with a TLS-encrypted web application on TCP port 443.
* Subnet *public-subnet* is a shared external subnet created by the cloud
operator which is reachable from the internet.
* We want to configure a basic load balancer that is accessible from the
internet, which distributes requests to the back-end servers.
* We want to employ a TCP health check to ensure that the back-end servers are
available.
**Solution**:
1. Create load balancer *lb1* on subnet *public-subnet*.
2. Create listener *listener1*.
3. Create pool *pool1* as *listener1*'s default pool.
4. Create a health monitor on *pool1* which probes *pool1*'s members' TCP
service port.
5. Add members 192.0.2.10 and 192.0.2.11 on *private-subnet* to *pool1*.
**CLI commands**:
::
neutron lbaas-loadbalancer-create --name lb1 public-subnet
# Re-run the following until lb1 shows ACTIVE and ONLINE statuses:
neutron lbaas-loadbalancer-show lb1
neutron lbaas-listener-create --name listener1 --loadbalancer lb1 --protocol HTTPS --protocol-port 443
neutron lbaas-pool-create --name pool1 --lb-algorithm ROUND_ROBIN --listener listener1 --protocol HTTPS
neutron lbaas-healthmonitor-create --delay 5 --max-retries 4 --timeout 10 --type TCP --pool pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.10 --protocol-port 443 pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.11 --protocol-port 443 pool1
.. _basic-tls-terminated-listener-neutron:
Deploy a TLS-terminated HTTPS load balancer
-------------------------------------------
With a TLS-terminated HTTPS load balancer, web clients communicate with the
load balancer over TLS protocols. The load balancer terminates the TLS session
and forwards the decrypted requests to the back-end servers. By terminating the
TLS session on the load balancer, we offload the CPU-intensive encryption work
to the load balancer, and enable the possibility of using advanced load
balancer features, like Layer 7 features and header manipulation.
**Scenario description**:
* Back-end servers 192.0.2.10 and 192.0.2.11 on subnet *private-subnet* have
been configured with regular HTTP application on TCP port 80.
* These back-end servers have been configured with a health check at the URL
path "/healthcheck". See :ref:`http-heath-monitors-neutron` below.
* Subnet *public-subnet* is a shared external subnet created by the cloud
operator which is reachable from the internet.
* A TLS certificate, key, and intermediate certificate chain for
www.example.com have been obtained from an external certificate authority.
These now exist in the files server.crt, server.key, and ca-chain.p7b in the
current directory. The key and certificate are PEM-encoded, and the
intermediate certificate chain is PKCS7 PEM encoded. The key is not encrypted
with a passphrase.
* The *admin* user on this cloud installation has keystone ID *admin_id*
* We want to configure a TLS-terminated HTTPS load balancer that is accessible
from the internet using the key and certificate mentioned above, which
distributes requests to the back-end servers over the non-encrypted HTTP
protocol.
**Solution**:
1. Create barbican *secret* resources for the certificate, key, and
intermediate certificate chain. We will call these *cert1*, *key1*, and
*intermediates1* respectively.
2. Create a *secret container* resource combining all of the above. We will
call this *tls_container1*.
3. Grant the *admin* user access to all the *secret* and *secret container*
barbican resources above.
4. Create load balancer *lb1* on subnet *public-subnet*.
5. Create listener *listener1* as a TERMINATED_HTTPS listener referencing
*tls_container1* as its default TLS container.
6. Create pool *pool1* as *listener1*'s default pool.
7. Add members 192.0.2.10 and 192.0.2.11 on *private-subnet* to *pool1*.
**CLI commands**:
::
openstack secret store --name='cert1' --payload-content-type='text/plain' --payload="$(cat server.crt)"
openstack secret store --name='key1' --payload-content-type='text/plain' --payload="$(cat server.key)"
openstack secret store --name='intermediates1' --payload-content-type='text/plain' --payload="$(cat ca-chain.p7b)"
openstack secret container create --name='tls_container1' --type='certificate' --secret="certificate=$(openstack secret list | awk '/ cert1 / {print $2}')" --secret="private_key=$(openstack secret list | awk '/ key1 / {print $2}')" --secret="intermediates=$(openstack secret list | awk '/ intermediates1 / {print $2}')"
neutron lbaas-loadbalancer-create --name lb1 public-subnet
# Re-run the following until lb1 shows ACTIVE and ONLINE statuses:
neutron lbaas-loadbalancer-show lb1
neutron lbaas-listener-create --loadbalancer lb1 --protocol-port 443 --protocol TERMINATED_HTTPS --name listener1 --default-tls-container=$(openstack secret container list | awk '/ tls_container1 / {print $2}')
neutron lbaas-pool-create --name pool1 --lb-algorithm ROUND_ROBIN --listener listener1 --protocol HTTP
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.10 --protocol-port 80 pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.11 --protocol-port 80 pool1
Deploy a TLS-terminated HTTPS load balancer with SNI
----------------------------------------------------
This example is exactly like :ref:`basic-tls-terminated-listener-neutron`,
except that we have multiple TLS certificates that we would like to use on
the same listener using Server Name Indication (SNI) technology.
**Scenario description**:
* Back-end servers 192.0.2.10 and 192.0.2.11 on subnet *private-subnet* have
been configured with regular HTTP application on TCP port 80.
* These back-end servers have been configured with a health check at the URL
path "/healthcheck". See :ref:`http-heath-monitors-neutron` below.
* Subnet *public-subnet* is a shared external subnet created by the cloud
operator which is reachable from the internet.
* TLS certificates, keys, and intermediate certificate chains for
www.example.com and www2.example.com have been obtained from an external
certificate authority. These now exist in the files server.crt, server.key,
ca-chain.p7b, server2.crt, server2-encrypted.key, and ca-chain2.p7b in the
current directory. The keys and certificates are PEM-encoded, and the
intermediate certificate chains are PKCS7 PEM encoded.
* The key for www.example.com is not encrypted with a passphrase.
* The key for www2.example.com is encrypted with the passphrase "abc123".
* The *admin* user on this cloud installation has keystone ID *admin_id*
* We want to configure a TLS-terminated HTTPS load balancer that is accessible
from the internet using the keys and certificates mentioned above, which
distributes requests to the back-end servers over the non-encrypted HTTP
protocol.
* If a web client connects that is not SNI capable, we want the load balancer
to respond with the certificate for www.example.com.
**Solution**:
1. Create barbican *secret* resources for the certificates, keys, and
intermediate certificate chains. We will call these *cert1*, *key1*,
*intermediates1*, *cert2*, *key2* and *intermediates2* respectively.
2. Create a barbican *secret* resource *passphrase2* for the passphrase for
*key2*
3. Create *secret container* resources combining the above appropriately. We
will call these *tls_container1* and *tls_container2*.
4. Grant the *admin* user access to all the *secret* and *secret container*
barbican resources above.
5. Create load balancer *lb1* on subnet *public-subnet*.
6. Create listener *listener1* as a TERMINATED_HTTPS listener referencing
*tls_container1* as its default TLS container, and referencing both
*tls_container1* and *tls_container2* using SNI.
7. Create pool *pool1* as *listener1*'s default pool.
8. Add members 192.0.2.10 and 192.0.2.11 on *private-subnet* to *pool1*.
**CLI commands**:
::
openstack secret store --name='cert1' --payload-content-type='text/plain' --payload="$(cat server.crt)"
openstack secret store --name='key1' --payload-content-type='text/plain' --payload="$(cat server.key)"
openstack secret store --name='intermediates1' --payload-content-type='text/plain' --payload="$(cat ca-chain.p7b)"
openstack secret container create --name='tls_container1' --type='certificate' --secret="certificate=$(openstack secret list | awk '/ cert1 / {print $2}')" --secret="private_key=$(openstack secret list | awk '/ key1 / {print $2}')" --secret="intermediates=$(openstack secret list | awk '/ intermediates1 / {print $2}')"
openstack secret store --name='cert2' --payload-content-type='text/plain' --payload="$(cat server2.crt)"
openstack secret store --name='key2' --payload-content-type='text/plain' --payload="$(cat server2-encrypted.key)"
openstack secret store --name='intermediates2' --payload-content-type='text/plain' --payload="$(cat ca-chain2.p7b)"
openstack secret store --name='passphrase2' --payload-content-type='text/plain' --payload="abc123"
openstack secret container create --name='tls_container2' --type='certificate' --secret="certificate=$(openstack secret list | awk '/ cert2 / {print $2}')" --secret="private_key=$(openstack secret list | awk '/ key2 / {print $2}')" --secret="intermediates=$(openstack secret list | awk '/ intermediates2 / {print $2}')" --secret="private_key_passphrase=$(openstack secret list | awk '/ passphrase2 / {print $2}')"
neutron lbaas-loadbalancer-create --name lb1 public-subnet
# Re-run the following until lb1 shows ACTIVE and ONLINE statuses:
neutron lbaas-loadbalancer-show lb1
neutron lbaas-listener-create --loadbalancer lb1 --protocol-port 443 --protocol TERMINATED_HTTPS --name listener1 --default-tls-container=$(openstack secret container list | awk '/ tls_container1 / {print $2}') --sni-container_refs $(openstack secret container list | awk '/ tls_container1 / {print $2}') $(openstack secret container list | awk '/ tls_container2 / {print $2}')
neutron lbaas-pool-create --name pool1 --lb-algorithm ROUND_ROBIN --listener listener1 --protocol HTTP
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.10 --protocol-port 80 pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.11 --protocol-port 80 pool1
Deploy HTTP and TLS-terminated HTTPS load balancing on the same IP and backend
------------------------------------------------------------------------------
This example is exactly like :ref:`basic-tls-terminated-listener-neutron`,
except that we would like to have both an HTTP and TERMINATED_HTTPS listener
that use the same back-end pool (and therefore, probably respond with the
exact same content regardless of whether the web client uses the HTTP or HTTPS
protocol to connect).
Please note that if you wish all HTTP requests to be redirected to HTTPS (so
that requests are only served via HTTPS, and attempts to access content over
HTTP just get redirected to the HTTPS listener), then please see `the example
<l7-cookbook-neutron.html#redirect-http-to-https-n>`__ in the
:doc:`l7-cookbook-neutron`.
**Scenario description**:
* Back-end servers 192.0.2.10 and 192.0.2.11 on subnet *private-subnet* have
been configured with regular HTTP application on TCP port 80.
* These back-end servers have been configured with a health check at the URL
path "/healthcheck". See :ref:`http-heath-monitors-neutron` below.
* Subnet *public-subnet* is a shared external subnet created by the cloud
operator which is reachable from the internet.
* A TLS certificate, key, and intermediate certificate chain for
www.example.com have been obtained from an external certificate authority.
These now exist in the files server.crt, server.key, and ca-chain.p7b in the
current directory. The key and certificate are PEM-encoded, and the
intermediate certificate chain is PKCS7 PEM encoded. The key is not encrypted
with a passphrase.
* The *admin* user on this cloud installation has keystone ID *admin_id*
* We want to configure a TLS-terminated HTTPS load balancer that is accessible
from the internet using the key and certificate mentioned above, which
distributes requests to the back-end servers over the non-encrypted HTTP
protocol.
* We also want to configure a HTTP load balancer on the same IP address as
the above which serves the exact same content (ie. forwards to the same
back-end pool) as the TERMINATED_HTTPS listener.
**Solution**:
1. Create barbican *secret* resources for the certificate, key, and
intermediate certificate chain. We will call these *cert1*, *key1*, and
*intermediates1* respectively.
2. Create a *secret container* resource combining all of the above. We will
call this *tls_container1*.
3. Grant the *admin* user access to all the *secret* and *secret container*
barbican resources above.
4. Create load balancer *lb1* on subnet *public-subnet*.
5. Create listener *listener1* as a TERMINATED_HTTPS listener referencing
*tls_container1* as its default TLS container.
6. Create pool *pool1* as *listener1*'s default pool.
7. Add members 192.0.2.10 and 192.0.2.11 on *private-subnet* to *pool1*.
8. Create listener *listener2* as an HTTP listener with *pool1* as its
default pool.
**CLI commands**:
::
openstack secret store --name='cert1' --payload-content-type='text/plain' --payload="$(cat server.crt)"
openstack secret store --name='key1' --payload-content-type='text/plain' --payload="$(cat server.key)"
openstack secret store --name='intermediates1' --payload-content-type='text/plain' --payload="$(cat ca-chain.p7b)"
openstack secret container create --name='tls_container1' --type='certificate' --secret="certificate=$(openstack secret list | awk '/ cert1 / {print $2}')" --secret="private_key=$(openstack secret list | awk '/ key1 / {print $2}')" --secret="intermediates=$(openstack secret list | awk '/ intermediates1 / {print $2}')"
neutron lbaas-loadbalancer-create --name lb1 public-subnet
# Re-run the following until lb1 shows ACTIVE and ONLINE statuses:
neutron lbaas-loadbalancer-show lb1
neutron lbaas-listener-create --loadbalancer lb1 --protocol-port 443 --protocol TERMINATED_HTTPS --name listener1 --default-tls-container=$(openstack secret container list | awk '/ tls_container1 / {print $2}')
neutron lbaas-pool-create --name pool1 --lb-algorithm ROUND_ROBIN --listener listener1 --protocol HTTP
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.10 --protocol-port 80 pool1
neutron lbaas-member-create --subnet private-subnet --address 192.0.2.11 --protocol-port 80 pool1
neutron lbaas-listener-create --name listener2 --loadbalancer lb1 --protocol HTTP --protocol-port 80 --default-pool pool1
.. _heath-monitor-best-practices-neutron:
Heath Monitor Best Practices
============================
While it is possible to set up a listener without a health monitor, if a
back-end pool member goes down, Octavia will not remove the failed server from
the pool until a considerable time has passed. This can lead to service
disruption for web clients. Because of this, we recommend always configuring
production load balancers to use a health monitor.
The health monitor itself is a process that does periodic health checks on each
back-end server to pre-emptively detect failed servers and temporarily pull
them out of the pool. Since effective health monitors depend as much on
back-end application server configuration as proper load balancer
configuration, some additional discussion of best practices is warranted here.
See also: `Octavia API Reference <https://developer.openstack.org/api-ref/load-balancer/>`_
Heath monitor options
---------------------
All of the health monitors Octavia supports have the following configurable
options:
* ``delay``: Number of seconds to wait between health checks.
* ``timeout``: Number of seconds to wait for any given health check to
complete. ``timeout`` should always be smaller than ``delay``.
* ``max-retries``: Number of subsequent health checks a given back-end
server must fail before it is considered *down*, or that a failed back-end
server must pass to be considered *up* again.
.. _http-heath-monitors-neutron:
HTTP health monitors
--------------------
In general, the application-side component of HTTP health checks are a part of
the web application being load balanced. By default, Octavia will probe the "/"
path on the application server. However, in many applications this is not
appropriate because the "/" path ends up being a cached page, or causes the
application server to do more work than is necessary for a basic health check.
In addition to the above options, HTTP health monitors also have the following
options:
* ``url_path``: Path part of the URL that should be retrieved from the back-end
server. By default this is "/".
* ``http_method``: HTTP method that should be used to retrieve the
``url_path``. By default this is "GET".
* ``expected_codes``: List of HTTP status codes that indicate an OK health
check. By default this is just "200".
Please keep the following best practices in mind when writing the code that
generates the health check in your web application:
* The health monitor ``url_path`` should not require authentication to load.
* By default the health monitor ``url_path`` should return a HTTP 200 OK status
code to indicate a healthy server unless you specify alternate
``expected_codes``.
* The health check should do enough internal checks to ensure the application
is healthy and no more. This may mean ensuring database or other external
storage connections are up and running, server load is acceptable, the site
is not in maintenance mode, and other tests specific to your application.
* The page generated by the health check should be very light weight:
* It should return in a sub-second interval.
* It should not induce significant load on the application server.
* The page generated by the health check should never be cached, though the
code running the health check may reference cached data. For example, you may
find it useful to run a more extensive health check via cron and store the
results of this to disk. The code generating the page at the health monitor
``url_path`` would incorporate the results of this cron job in the tests it
performs.
* Since Octavia only cares about the HTTP status code returned, and since
health checks are run so frequently, it may make sense to use the "HEAD" or
"OPTIONS" HTTP methods to cut down on unnecessary processing of a whole page.
Other heath monitors
--------------------
Other health monitor types include ``PING``, ``TCP``, ``HTTPS``, and
``TLS-HELLO``.
``PING`` health monitors send periodic ICMP PING requests to the back-end
servers. Obviously, your back-end servers must be configured to allow PINGs in
order for these health checks to pass.
``TCP`` health monitors open a TCP connection to the back-end server's protocol
port. Your custom TCP application should be written to respond OK to the load
balancer connecting, opening a TCP connection, and closing it again after the
TCP handshake without sending any data.
``HTTPS`` health monitors operate exactly like HTTP health monitors, but with
ssl back-end servers. Unfortunately, this causes problems if the servers are
performing client certificate validation, as HAProxy won't have a valid cert.
In this case, using ``TLS-HELLO`` type monitoring is an alternative.
``TLS-HELLO`` health monitors simply ensure the back-end server responds to
SSLv3 client hello messages. It will not check any other health metrics, like
status code or body contents.
Intermediate certificate chains
===============================
Some TLS certificates require you to install an intermediate certificate chain
in order for web client browsers to trust the certificate. This chain can take
several forms, and is a file provided by the organization from whom you
obtained your TLS certificate.
PEM-encoded chains
------------------
The simplest form of the intermediate chain is a PEM-encoded text file that
either contains a sequence of individually-encoded PEM certificates, or a PEM
encoded PKCS7 block(s). If this is the type of intermediate chain you have been
provided, the file will contain either ``-----BEGIN PKCS7-----`` or
``-----BEGIN CERTIFICATE-----`` near the top of the file, and one or more
blocks of 64-character lines of ASCII text (that will look like gobbedlygook to
a human). These files are also typically named with a ``.crt`` or ``.pem``
extension.
To upload this type of intermediates chain to barbican, run a command similar
to the following (assuming "intermediates-chain.pem" is the name of the file):
::
openstack secret store --name='intermediates1' --payload-content-type='text/plain' --payload="$(cat intermediates-chain.pem)"
DER-encoded chains
------------------
If the intermediates chain provided to you is a file that contains what appears
to be random binary data, it is likely that it is a PKCS7 chain in DER format.
These files also may be named with a ``.p7b`` extension. In order to use this
intermediates chain, you can either convert it to a series of PEM-encoded
certificates with the following command:
::
openssl pkcs7 -in intermediates-chain.p7b -inform DER -print_certs -out intermediates-chain.pem
...or convert it into a PEM-encoded PKCS7 bundle with the following command:
::
openssl pkcs7 -in intermediates-chain.p7b -inform DER -outform PEM -out intermediates-chain.pem
...or simply upload the binary DER file to barbican without conversion:
::
openstack secret store --name='intermediates1' --payload-content-type='application/octet-stream' --payload-content-encoding='base64' --payload="$(cat intermediates-chain.p7b | base64)"
In any case, if the file is not a PKCS7 DER bundle, then either of the above
two openssl commands will fail.
Further reading
===============
For examples of using Layer 7 features for more advanced load balancing, please
see: :doc:`l7-cookbook-neutron`

361
doc/source/user/guides/l7-cookbook-neutron.rst
View File

@ -1,361 +0,0 @@
..
Copyright (c) 2016 IBM
Licensed under the Apache License, Version 2.0 (the "License"); you may
not use this file except in compliance with the License. You may obtain
a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
License for the specific language governing permissions and limitations
under the License.
==================================================
Layer 7 Cookbook Using Neutron Client (deprecated)
==================================================
.. warning:: The neutron client used in this document is deprecated. We
strongly encourage you to use the OpenStack Client and Octavia
OpenStack Client plugin instead. This document is being maintained
for deployments still using neutron-lbaas and the neutron client.
Introduction
============
This document gives several examples of common L7 load balancer usage. For a
description of L7 load balancing see: :doc:`l7`
For the purposes of this guide we assume that the neutron command-line
interface is going to be used to configure all features of Neutron LBaaS with
an Octavia back-end. Also, in order to keep these examples short, we assume
that many non-L7 configuration tasks (such as deploying loadbalancers,
listeners, pools, members, healthmonitors, etc.) have already been
accomplished. A description of the starting conditions is given in each example
below.
Examples
========
.. _redirect-http-to-https-n:
Redirect *http://www.example.com/* to *https://www.example.com/*
----------------------------------------------------------------
**Scenario description**:
* Load balancer *lb1* has been set up with ``TERMINATED_HTTPS`` listener
*tls_listener* on TCP port 443.
* *tls_listener* has been populated with a default pool, members, etc.
* *tls_listener* is available under the DNS name *https://www.example.com/*
* We want any regular HTTP requests to TCP port 80 on *lb1* to be redirected
to *tls_listener* on TCP port 443.
**Solution**:
1. Create listener *http_listener* as an HTTP listener on *lb1* port 80.
2. Set up an L7 Policy *policy1* on *http_listener* with action
``REDIRECT_TO_URL`` pointed at the URL *https://www.example.com/*
3. Add an L7 Rule to *policy1* which matches all requests.
**CLI commands**:
::
neutron lbaas-listener-create --name http_listener --loadbalancer lb1 --protocol HTTP --protocol-port 80
neutron lbaas-l7policy-create --action REDIRECT_TO_URL --redirect-url https://www.example.com/ --listener http_listener --name policy1
neutron lbaas-l7rule-create --type PATH --compare-type STARTS_WITH --value / policy1
.. _send-requests-to-static-pool-n:
Send requests starting with /js or /images to *static_pool*
-----------------------------------------------------------
**Scenario description**:
* Listener *listener1* on load balancer *lb1* is set up to send all requests to
its default_pool *pool1*.
* We are introducing static content servers 10.0.0.10 and 10.0.0.11 on subnet
*private-subnet*, and want any HTTP requests with a URL that starts with
either "/js" or "/images" to be sent to those two servers instead of *pool1*.
**Solution**:
1. Create pool *static_pool* on *lb1*.
2. Populate *static_pool* with the new back-end members.
3. Create L7 Policy *policy1* with action ``REDIRECT_TO_POOL`` pointed at
*static_pool*.
4. Create an L7 Rule on *policy1* which looks for "/js" at the start of
the request path.
5. Create L7 Policy *policy2* with action ``REDIRECT_TO_POOL`` pointed at
*static_pool*.
6. Create an L7 Rule on *policy2* which looks for "/images" at the start
of the request path.
**CLI commands**:
::
neutron lbaas-pool-create --name static_pool --lb-algorithm ROUND_ROBIN --loadbalancer lb1 --protocol HTTP
neutron lbaas-member-create --subnet private-subnet --address 10.0.0.10 --protocol-port 80 static_pool
neutron lbaas-member-create --subnet private-subnet --address 10.0.0.11 --protocol-port 80 static_pool
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool static_pool --listener listener1 --name policy1
neutron lbaas-l7rule-create --type PATH --compare-type STARTS_WITH --value /js policy1
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool static_pool --listener listener1 --name policy2
neutron lbaas-l7rule-create --type PATH --compare-type STARTS_WITH --value /images policy2
**Alternate solution** (using regular expressions):
1. Create pool *static_pool* on *lb1*.
2. Populate *static_pool* with the new back-end members.
3. Create L7 Policy *policy1* with action ``REDIRECT_TO_POOL`` pointed at
*static_pool*.
4. Create an L7 Rule on *policy1* which uses a regular expression to match
either "/js" or "/images" at the start of the request path.
**CLI commands**:
::
neutron lbaas-pool-create --name static_pool --lb-algorithm ROUND_ROBIN --loadbalancer lb1 --protocol HTTP
neutron lbaas-member-create --subnet private-subnet --address 10.0.0.10 --protocol-port 80 static_pool
neutron lbaas-member-create --subnet private-subnet --address 10.0.0.11 --protocol-port 80 static_pool
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool static_pool --listener listener1 --name policy1
neutron lbaas-l7rule-create --type PATH --compare-type REGEX --value '^/(js|images)' policy1
Send requests for *http://www2.example.com/* to *pool2*
-------------------------------------------------------
**Scenario description**:
* Listener *listener1* on load balancer *lb1* is set up to send all requests to
its default_pool *pool1*.
* We have set up a new pool *pool2* on *lb1* and want any requests using the
HTTP/1.1 hostname *www2.example.com* to be sent to *pool2* instead.
**Solution**:
1. Create L7 Policy *policy1* with action ``REDIRECT_TO_POOL`` pointed at
*pool2*.
2. Create an L7 Rule on *policy1* which matches the hostname
*www2.example.com*.
**CLI commands**:
::
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool pool2 --listener listener1 --name policy1
neutron lbaas-l7rule-create --type HOST_NAME --compare-type EQUAL_TO --value www2.example.com policy1
Send requests for *\*.example.com* to *pool2*
---------------------------------------------
**Scenario description**:
* Listener *listener1* on load balancer *lb1* is set up to send all requests to
its default_pool *pool1*.
* We have set up a new pool *pool2* on *lb1* and want any requests using any
HTTP/1.1 hostname like *\*.example.com* to be sent to *pool2* instead.
**Solution**:
1. Create L7 Policy *policy1* with action ``REDIRECT_TO_POOL`` pointed at
*pool2*.
2. Create an L7 Rule on *policy1* which matches any hostname that ends with
*example.com*.
**CLI commands**:
::
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool pool2 --listener listener1 --name policy1
neutron lbaas-l7rule-create --type HOST_NAME --compare-type ENDS_WITH --value example.com policy1
Send unauthenticated users to *login_pool* (scenario 1)
-------------------------------------------------------
**Scenario description**:
* ``TERMINATED_HTTPS`` listener *listener1* on load balancer *lb1* is set up
to send all requests to its default_pool *pool1*.
* The site behind *listener1* requires all web users to authenticate, after
which a browser cookie *auth_token* will be set.
* When web users log out, or if the *auth_token* is invalid, the application
servers in *pool1* clear the *auth_token*.
* We want to introduce new secure authentication server 10.0.1.10 on Neutron
subnet *secure_subnet* (a different Neutron subnet from the default
application servers) which handles authenticating web users and sets the
*auth_token*.
*Note:* Obviously, to have a more secure authentication system that is less
vulnerable to attacks like XSS, the new secure authentication server will need
to set session variables to which the default_pool servers will have access
outside the data path with the web client. There may be other security concerns
as well. This example is not meant to address how these are to be
accomplished--it's mainly meant to show how L7 application routing can be done
based on a browser cookie.
**Solution**:
1. Create pool *login_pool* on *lb1*.
2. Add member 10.0.1.10 on *secure_subnet* to *login_pool*.
3. Create L7 Policy *policy1* with action ``REDIRECT_TO_POOL`` pointed at
*login_pool*.
4. Create an L7 Rule on *policy1* which looks for browser cookie *auth_token*
(with any value) and matches if it is *NOT* present.
**CLI commands**:
::
neutron lbaas-pool-create --name login_pool --lb-algorithm ROUND_ROBIN --loadbalancer lb1 --protocol HTTP
neutron lbaas-member-create --subnet secure_subnet --address 10.0.1.10 --protocol-port 80 login_pool
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool login_pool --listener listener1 --name policy1
neutron lbaas-l7rule-create --type COOKIE --key auth_token --compare-type REGEX --value '.*' --invert policy1
Send unauthenticated users to *login_pool* (scenario 2)
--------------------------------------------------------
**Scenario description**:
* ``TERMINATED_HTTPS`` listener *listener1* on load balancer *lb1* is set up
to send all requests to its default_pool *pool1*.
* The site behind *listener1* requires all web users to authenticate, after
which a browser cookie *auth_token* will be set.
* When web users log out, or if the *auth_token* is invalid, the application
servers in *pool1* set *auth_token* to the literal string "INVALID".
* We want to introduce new secure authentication server 10.0.1.10 on Neutron
subnet *secure_subnet* (a different Neutron subnet from the default
application servers) which handles authenticating web users and sets the
*auth_token*.
*Note:* Obviously, to have a more secure authentication system that is less
vulnerable to attacks like XSS, the new secure authentication server will need
to set session variables to which the default_pool servers will have access
outside the data path with the web client. There may be other security concerns
as well. This example is not meant to address how these are to be
accomplished-- it's mainly meant to show how L7 application routing can be done
based on a browser cookie.
**Solution**:
1. Create pool *login_pool* on *lb1*.
2. Add member 10.0.1.10 on *secure_subnet* to *login_pool*.
3. Create L7 Policy *policy1* with action ``REDIRECT_TO_POOL`` pointed at
*login_pool*.
4. Create an L7 Rule on *policy1* which looks for browser cookie *auth_token*
(with any value) and matches if it is *NOT* present.
5. Create L7 Policy *policy2* with action ``REDIRECT_TO_POOL`` pointed at
*login_pool*.
6. Create an L7 Rule on *policy2* which looks for browser cookie *auth_token*
and matches if it is equal to the literal string "INVALID".
**CLI commands**:
::
neutron lbaas-pool-create --name login_pool --lb-algorithm ROUND_ROBIN --loadbalancer lb1 --protocol HTTP
neutron lbaas-member-create --subnet secure_subnet --address 10.0.1.10 --protocol-port 80 login_pool
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool login_pool --listener listener1 --name policy1
neutron lbaas-l7rule-create --type COOKIE --key auth_token --compare-type REGEX --value '.*' --invert policy1
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool login_pool --listener listener1 --name policy2
neutron lbaas-l7rule-create --type COOKIE --key auth_token --compare-type EQUAL_TO --value INVALID policy2
Send requests for *http://api.example.com/api* to *api_pool*
------------------------------------------------------------
**Scenario description**:
* Listener *listener1* on load balancer *lb1* is set up to send all requests
to its default_pool *pool1*.
* We have created pool *api_pool* on *lb1*, however, for legacy business logic
reasons, we only want requests sent to this pool if they match the hostname
*api.example.com* AND the request path starts with */api*.
**Solution**:
1. Create L7 Policy *policy1* with action ``REDIRECT_TO_POOL`` pointed at
*api_pool*.
2. Create an L7 Rule on *policy1* which matches the hostname *api.example.com*.
3. Create an L7 Rule on *policy1* which matches */api* at the start of the
request path. (This rule will be logically ANDed with the previous rule.)
**CLI commands**:
::
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool api_pool --listener listener1 --name policy1
neutron lbaas-l7rule-create --type HOST_NAME --compare-type EQUAL_TO --value api.example.com policy1
neutron lbaas-l7rule-create --type PATH --compare-type STARTS_WITH --value /api policy1
Set up A/B testing on an existing production site using a cookie
----------------------------------------------------------------
**Scenario description**:
* Listener *listener1* on load balancer *lb1* is a production site set up as
described under :ref:`send-requests-to-static-pool-n` (alternate solution)
above. Specifically:
* HTTP requests with a URL that starts with either "/js" or "/images" are
sent to pool *static_pool*.
* All other requests are sent to *listener1's* default_pool *pool1*.
* We are introducing a "B" version of the production site, complete with its
own default_pool and static_pool. We will call these *pool_B* and
*static_pool_B* respectively.
* The *pool_B* members should be 10.0.0.50 and 10.0.0.51, and the
*static_pool_B* members should be 10.0.0.100 and 10.0.0.101 on subnet
*private-subnet*.
* Web clients which should be routed to the "B" version of the site get a
cookie set by the member servers in *pool1*. This cookie is called
"site_version" and should have the value "B".
**Solution**:
1. Create pool *pool_B* on *lb1*.
2. Populate *pool_B* with its new back-end members.
3. Create pool *static_pool_B* on *lb1*.
4. Populate *static_pool_B* with its new back-end members.
5. Create L7 Policy *policy2* with action ``REDIRECT_TO_POOL`` pointed at
*static_pool_B*. This should be inserted at position 1.
6. Create an L7 Rule on *policy2* which uses a regular expression to match
either "/js" or "/images" at the start of the request path.
7. Create an L7 Rule on *policy2* which matches the cookie "site_version" to
the exact string "B".
8. Create L7 Policy *policy3* with action ``REDIRECT_TO_POOL`` pointed at
*pool_B*. This should be inserted at position 2.
9. Create an L7 Rule on *policy3* which matches the cookie "site_version" to
the exact string "B".
*A word about L7 Policy position*: Since L7 Policies are evaluated in order
according to their position parameter, and since the first L7 Policy whose L7
Rules all evaluate to True is the one whose action is followed, it is important
that L7 Policies with the most specific rules get evaluated first.
For example, in this solution, if *policy3* were to appear in the listener's L7
Policy list before *policy2* (that is, if *policy3* were to have a lower
position number than *policy2*), then if a web client were to request the URL
http://www.example.com/images/a.jpg with the cookie "site_version:B", then
*policy3* would match, and the load balancer would send the request to
*pool_B*. From the scenario description, this request clearly was meant to be
sent to *static_pool_B*, which is why *policy2* needs to be evaluated before
*policy3*.
**CLI commands**:
::
neutron lbaas-pool-create --name pool_B --lb-algorithm ROUND_ROBIN --loadbalancer lb1 --protocol HTTP
neutron lbaas-member-create --subnet private-subnet --address 10.0.0.50 --protocol-port 80 pool_B
neutron lbaas-member-create --subnet private-subnet --address 10.0.0.51 --protocol-port 80 pool_B
neutron lbaas-pool-create --name static_pool_B --lb-algorithm ROUND_ROBIN --loadbalancer lb1 --protocol HTTP
neutron lbaas-member-create --subnet private-subnet --address 10.0.0.100 --protocol-port 80 static_pool_B
neutron lbaas-member-create --subnet private-subnet --address 10.0.0.101 --protocol-port 80 static_pool_B
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool static_pool_B --listener listener1 --name policy2 --position 1
neutron lbaas-l7rule-create --type PATH --compare-type REGEX --value '^/(js|images)' policy2
neutron lbaas-l7rule-create --type COOKIE --key site_version --compare-type EQUAL_TO --value B policy2
neutron lbaas-l7policy-create --action REDIRECT_TO_POOL --redirect-pool pool_B --listener listener1 --name policy3 --position 2
neutron lbaas-l7rule-create --type COOKIE --key site_version --compare-type EQUAL_TO --value B policy3

2
doc/source/user/index.rst
View File

@ -10,8 +10,6 @@ Cookbooks
guides/basic-cookbook
guides/l7-cookbook
guides/basic-cookbook-neutron
guides/l7-cookbook-neutron
Guides
======

24
etc/octavia.conf
View File

@ -19,7 +19,6 @@
[api_settings]
# bind_host = 127.0.0.1
# bind_port = 9876
# api_handler = queue_producer
# How should authentication be handled (keystone, noauth)
# auth_strategy = keystone
@ -33,10 +32,6 @@
# api_base_uri = http://localhost:9876
# api_base_uri =
# Enable/disable exposing API endpoints. By default, both v1 and v2 are enabled.
# api_v1_enabled = True
# api_v2_enabled = True
# Enable/disable ability for users to create TLS Terminated listeners
# allow_tls_terminated_listeners = True
@ -95,14 +90,6 @@
# health_update_driver = health_db
# stats_update_driver = stats_db
# EventStreamer options are
# queue_event_streamer,
# noop_event_streamer
# event_streamer_driver = noop_event_streamer
# Enable provisioning status sync with neutron db
# sync_provisioning_status = False
[keystone_authtoken]
# This group of config options are imported from keystone middleware. Thus the
# option names should match the names declared in the middleware.
@ -285,17 +272,6 @@
# Topic (i.e. Queue) Name
# topic = octavia_prov
# Topic for octavia's events sent to a queue
# event_stream_topic = neutron_lbaas_event
# Transport URL to use for the neutron-lbaas synchronization event stream
# when neutron and octavia have separate queues.
# For Single Host, specify one full transport URL:
# event_stream_transport_url = rabbit://<user>:<pass>@127.0.0.1:5672/<vhost>
# For HA, specify queue nodes in cluster, comma delimited:
# event_stream_transport_url = rabbit://<user>:<pass>@server01,<user>:<pass>@server02/<vhost>
# event_stream_transport_url =
[house_keeping]
# Interval in seconds to initiate spare amphora checks
# spare_check_interval = 30

11
octavia/api/handlers/__init__.py
View File

@ -1,11 +0,0 @@
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.

67
octavia/api/handlers/abstract_handler.py
View File

@ -1,67 +0,0 @@
# Copyright 2014 Rackspace
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
import abc
import six
@six.add_metaclass(abc.ABCMeta)
class BaseObjectHandler(object):
"""Base class for any object handler."""
@abc.abstractmethod
def create(self, model_id):
"""Begins process of actually creating data_model."""
pass
@abc.abstractmethod
def update(self, model_id, updated_dict):
"""Begins process of actually updating data_model."""
pass
@abc.abstractmethod
def delete(self, model_id):
"""Begins process of actually deleting data_model."""
pass
class NotImplementedObjectHandler(BaseObjectHandler):
"""Default Object Handler to force implementation of subclasses.
Helper class to make any subclass of AbstractHandler explode if it
is missing any of the required object managers.
"""
@staticmethod
def update(model_id, updated_dict):
raise NotImplementedError()
@staticmethod
def delete(model_id):
raise NotImplementedError()
@staticmethod
def create(model_id):
raise NotImplementedError()
@six.add_metaclass(abc.ABCMeta)
class BaseHandler(object):
"""Base class for all handlers."""
load_balancer = NotImplementedObjectHandler()
listener = NotImplementedObjectHandler()
pool = NotImplementedObjectHandler()
health_monitor = NotImplementedObjectHandler()
member = NotImplementedObjectHandler()
l7policy = NotImplementedObjectHandler()
l7rule = NotImplementedObjectHandler()

11
octavia/api/handlers/controller_simulator/__init__.py
View File

@ -1,11 +0,0 @@
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.

478
octavia/api/handlers/controller_simulator/handler.py
View File

@ -1,478 +0,0 @@
# Copyright 2014 Rackspace
# Copyright 2016 Blue Box, an IBM Company
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
"""
This is just a handler that will simulate successful operations a controller
should perform. There is nothing useful about this other than database
entity status management.
"""
import threading
import time
from oslo_log import log as logging
from octavia.api.handlers import abstract_handler
from octavia.common import constants
from octavia.common import data_models
from octavia.db import api as db_api
import octavia.db.repositories as repos
LOG = logging.getLogger(__name__)
ASYNC_TIME = 1
def validate_input(expected, actual):
if not isinstance(actual, expected):
raise InvalidHandlerInputObject(obj_type=actual.__class__)
def simulate_controller(data_model, delete=False, update=False, create=False,
batch_update=False):
"""Simulates a successful controller operator for a data model.
:param data_model: data model to simulate controller operation
:param delete: deletes from the database
"""
repo = repos.Repositories()
def member_controller(member, delete=False, update=False, create=False,
batch_update=False):
time.sleep(ASYNC_TIME)
LOG.info("Simulating controller operation for member...")
db_mem = None
if delete:
db_mem = repo.member.get(db_api.get_session(), id=member.id)
repo.member.delete(db_api.get_session(), id=member.id)
elif update:
db_mem = repo.member.get(db_api.get_session(), id=member.id)
member_dict = member.to_dict()
member_dict['operating_status'] = db_mem.operating_status
repo.member.update(db_api.get_session(), member.id, **member_dict)
elif create:
repo.member.update(db_api.get_session(), member.id,
operating_status=constants.ONLINE)
elif batch_update:
members = member
for m in members:
repo.member.update(db_api.get_session(), m.id,
operating_status=constants.ONLINE)
listeners = []
if db_mem:
for listener in db_mem.pool.listeners:
if listener not in listeners:
listeners.append(listener)
if member.pool.listeners:
for listener in member.pool.listeners:
if listener not in listeners:
listeners.append(listener)
if listeners:
for listener in listeners:
repo.listener.update(db_api.get_session(), listener.id,
operating_status=constants.ONLINE,
provisioning_status=constants.ACTIVE)
repo.load_balancer.update(db_api.get_session(),
member.pool.load_balancer.id,
operating_status=constants.ONLINE,
provisioning_status=constants.ACTIVE)
LOG.info("Simulated Controller Handler Thread Complete")
def l7policy_controller(l7policy, delete=False, update=False,
create=False):
time.sleep(ASYNC_TIME)
LOG.info("Simulating controller operation for l7policy...")
db_l7policy = None
if delete:
db_l7policy = repo.l7policy.get(db_api.get_session(),
id=l7policy.id)
repo.l7policy.delete(db_api.get_session(), id=l7policy.id)
elif update:
db_l7policy = repo.l7policy.get(db_api.get_session(),
id=l7policy.id)
l7policy_dict = l7policy.to_dict()
repo.l7policy.update(db_api.get_session(), l7policy.id,
**l7policy_dict)
elif create:
db_l7policy = repo.l7policy.create(db_api.get_session(),
**l7policy.to_dict())
if db_l7policy.listener:
repo.listener.update(db_api.get_session(), db_l7policy.listener.id,
operating_status=constants.ONLINE,
provisioning_status=constants.ACTIVE)
repo.load_balancer.update(db_api.get_session(),
db_l7policy.listener.load_balancer.id,
operating_status=constants.ONLINE,
provisioning_status=constants.ACTIVE)
LOG.info("Simulated Controller Handler Thread Complete")
def l7rule_controller(l7rule, delete=False, update=False, create=False):
time.sleep(ASYNC_TIME)
LOG.info("Simulating controller operation for l7rule...")
db_l7rule = None
if delete:
db_l7rule = repo.l7rule.get(db_api.get_session(), id=l7rule.id)
repo.l7rule.delete(db_api.get_session(), id=l7rule.id)
elif update:
db_l7rule = repo.l7rule.get(db_api.get_session(), id=l7rule.id)
l7rule_dict = l7rule.to_dict()
repo.l7rule.update(db_api.get_session(), l7rule.id, **l7rule_dict)
elif create:
l7rule_dict = l7rule.to_dict()
db_l7rule = repo.l7rule.create(db_api.get_session(), **l7rule_dict)
if db_l7rule.l7policy.listener:
listener = db_l7rule.l7policy.listener
repo.listener.update(db_api.get_session(), listener.id,
operating_status=constants.ONLINE,
provisioning_status=constants.ACTIVE)
repo.load_balancer.update(db_api.get_session(),
listener.load_balancer.id,
operating_status=constants.ONLINE,
provisioning_status=constants.ACTIVE)
LOG.info("Simulated Controller Handler Thread Complete")
def health_monitor_controller(health_monitor, delete=False, update=False,
create=False):
time.sleep(ASYNC_TIME)
LOG.info("Simulating controller operation for health monitor...")
db_hm = None
if delete:
db_hm = repo.health_monitor.get(db_api.get_session(),
pool_id=health_monitor.pool.id)
repo.health_monitor.delete(db_api.get_session(),
pool_id=health_monitor.pool.id)
elif update:
db_hm = repo.health_monitor.get(db_api.get_session(),
pool_id=health_monitor.pool_id)
hm_dict = health_monitor.to_dict()
hm_dict['operating_status'] = db_hm.operating_status()
repo.health_monitor.update(db_api.get_session(), **hm_dict)
elif create:
repo.pool.update(db_api.get_session(), health_monitor.pool_id,
operating_status=constants.ONLINE)
listeners = []
if db_hm:
for listener in db_hm.pool.listeners:
if listener not in listeners:
listeners.append(listener)
if health_monitor.pool.listeners:
for listener in health_monitor.pool.listeners:
if listener not in listeners:
listeners.append(listener)
if listeners:
for listener in listeners:
repo.test_and_set_lb_and_listener_prov_status(
db_api.get_session(),
health_monitor.pool.load_balancer.id,
listener.id, constants.ACTIVE,
constants.ACTIVE)
repo.listener.update(db_api.get_session(),
listener.id,
operating_status=constants.ONLINE,
provisioning_status=constants.ACTIVE)
repo.load_balancer.update(
db_api.get_session(),
health_monitor.pool.load_balancer.id,
operating_status=constants.ONLINE,
provisioning_status=constants.ACTIVE)
LOG.info("Simulated Controller Handler Thread Complete")
def pool_controller(pool, delete=False, update=False, create=False):
time.sleep(ASYNC_TIME)
LOG.info("Simulating controller operation for pool...")
db_pool = None
if delete:
db_pool = repo.pool.get(db_api.get_session(), id=pool.id)
repo.pool.delete(db_api.get_session(), id=pool.id)
elif update:
db_pool = repo.pool.get(db_api.get_session(), id=pool.id