7b3483723a
Changes to get jobs working 1) After [1] devstack no longer changes the ownership of the whole /opt/stack tree to the stack user unconditionally. Switch to the stack user when running integration test. 2) Add bindep.txt file[2]. The default fallback file is not installed anymore and therefore a bindep.txt file is needed to add install additional packages. 3) Use trovestack script rather than devstack to build image so many global variables could be used for consistency. By default, devstack won't build image. 4) Remove the tools/test-setup.sh as it is not used any more. 5) Instance upgrade test keeps failing in CI for some reason, although it's always passed on my local environment. In order not to block other patches, skip the instance upgrade tests temporarily. [1] https://review.opendev.org/203698 [2] http://lists.openstack.org/pipermail/openstack-discuss/2019-June/006888.html Change-Id: I35e17afb9e827b1fead9d28dbf32f11ce4034a9b |
||
---|---|---|
.. | ||
scripts | ||
tests/integration | ||
README.md |
Trove integration script - trovestack
Steps to setup environment
Install a fresh Ubuntu 16.04 (xenial) image. We suggest creating a development virtual machine using the image.
-
Login to the machine as root
-
Make sure we have git installed
# apt-get update # apt-get install git-core -y
-
Add a user named ubuntu if you do not already have one:
# adduser ubuntu # visudo
Add this line to the file below the root user
ubuntu ALL=(ALL:ALL) ALL
Or use this if you dont want to type your password to sudo a command:
ubuntu ALL=(ALL) NOPASSWD: ALL
if /dev/pts/0 does not have read/write for your user
# chmod 666 /dev/pts/0
Note that this number can change and if you can not connect to the screen session then the /dev/pts/# needs modding like above.
-
Login with ubuntu and download the Trove code.
# su ubuntu $ mkdir -p /opt/stack $ cd /opt/stack
Note that it is important that you clone the repository here. This is a change from the earlier trove-integration where you could clone trove-integration anywhere you wanted (like HOME) and trove would get cloned for you in the right place. Since trovestack is now in the trove repository, if you wish to test changes that you have made to trove, it is advisable for you to have your trove repository in /opt/stack to avoid another trove repository being cloned for you.
-
Clone this repo and go into the scripts directory
$ git clone https://github.com/openstack/trove.git $ cd trove/integration/scripts/
Running trovestack
Run this to get the command list with a short description of each
$ ./trovestack
Install Trove
This brings up trove services and initializes the trove database.
$ ./trovestack install
Connecting to the screen session
$ screen -x stack
If that command fails with the error
Cannot open your terminal '/dev/pts/1'
If that command fails with the error chmod the corresponding /dev/pts/#
$ chmod 660 /dev/pts/1
Navigate the log screens
To produce the list of screens that you can scroll through and select
ctrl+a then "
An example of screen list:
..... (full list ommitted)
20 c-vol
21 h-eng
22 h-api
23 h-api-cfn
24 h-api-cw
25 tr-api
26 tr-tmgr
27 tr-cond
Alternatively, to go directly to a specific screen window
ctrl+a then '
then enter a number (like 25) or name (like tr-api)
Detach from the screen session
Allows the services to continue running in the background
ctrl+a then d
Kick start the build/test-init/build-image commands
Add mysql as a parameter to set build and add the mysql guest image. This will also populate /etc/trove/test.conf with appropriate values for running the integration tests.
$ ./trovestack kick-start mysql
Initialize the test configuration and set up test users (overwrites /etc/trove/test.conf)
$ ./trovestack test-init
Build guest agent image
The trove guest agent image could be created using trovestack
script
according to the following command:
PATH_DEVSTACK_OUTPUT=/opt/stack \
./trovestack build-image \
${datastore_type} \
${guest_os} \
${guest_os_release} \
${dev_mode}
- If the script is running as a part of DevStack, the viriable
PATH_DEVSTACK_OUTPUT
is set automatically. - if
dev_mode=false
, the trove code for guest agent is injected into the image at the building time. Nowdev_mode=false
is still in experimental and not considered production ready yet. - If
dev_mode=true
, the script assumes to be running on the host of trove-taskmanager, otherwise,CONTROLLER_IP
needs to be specified explicitly.
For example, build a Mysql image for Ubuntu Xenial operating system:
$ ./trovestack build-image mysql ubuntu xenial false
Running Integration Tests
Check the values in /etc/trove/test.conf in case it has been re-initialized prior to running the tests. For example, from the previous mysql steps:
"dbaas_datastore": "%datastore_type%",
"dbaas_datastore_version": "%datastore_version%",
should be:
"dbaas_datastore": "mysql",
"dbaas_datastore_version": "5.5",
Once Trove is running on DevStack, you can run the integration tests locally.
$./trovestack int-tests
This will runs all of the blackbox tests by default. Use the --group
option to run a different group:
$./trovestack int-tests --group=simple_blackbox
You can also specify the TESTS_USE_INSTANCE_ID
environment variable to have the integration tests use an existing instance for the tests rather than creating a new one.
$./TESTS_DO_NOT_DELETE_INSTANCE=True TESTS_USE_INSTANCE_ID=INSTANCE_UUID ./trovestack int-tests --group=simple_blackbox
Reset your environment
Stop all the services running in the screens and refresh the environment
$ killall -9 screen
$ screen -wipe
$ RECLONE=yes ./trovestack install
$ ./trovestack kick-start mysql
or
$ RECLONE=yes ./trovestack install
$ ./trovestack test-init
$ ./trovestack build-image mysql
Recover after reboot
If the VM was restarted, then the process for bringing up Openstack and Trove is quite simple
$./trovestack start-deps
$./trovestack start
Use screen to ensure all modules have started without error
$screen -r stack
VMware Fusion 5 speed improvement
Running Ubuntu with KVM or Qemu can be extremely slow without certain optimizations. The following are some VMware settings that can improve performance and may also apply to other virtualization platforms.
-
Shutdown the Ubuntu VM.
-
Go to VM Settings -> Processors & Memory -> Advanced Options. Check the "Enable hypervisor applications in this virtual machine"
-
Go to VM Settings -> Advanced. Set the "Troubleshooting" option to "None"
-
After setting these create a snapshot so that in cases where things break down you can revert to a clean snapshot.
-
Boot up the VM and run the
./trovestack install
-
To verify that KVM is setup properly after the devstack installation you can run these commands.
ubuntu@ubuntu:~$ kvm-ok
INFO: /dev/kvm exists
KVM acceleration can be used
VMware Workstation performance improvements
In recent versions of VMWare, you can get much better performance if you enable the right virtualization options. For example, in VMWare Workstation (found in version 10.0.2), click on VM->Settings->Processor.
You should see a box of "Virtualization Engine" options that can be changed only when the VM is shutdown.
Make sure you check "Virtualize Intel VT-x/EPT or AMD-V/RVI" and "Virtualize CPU performance counters". Set the preferred mode to "Automatic".
Then boot the VM and ensure that the proper virtualization is enabled.
ubuntu@ubuntu:~$ kvm-ok
INFO: /dev/kvm exists
KVM acceleration can be used