Major issues encountered when debugging OpenStack (by quqi99)

Zhang Hua posted: 2014-11-09
Copyright notice: Can be reproduced arbitrarily, please be sure to use hyperlinks in the form of the original source of the article and the author's information and this copyright notice (http://blog.csdn.net/quqi99)

Today want to debug the Nova-compute process, with devstack rapid installation, kill off the Nova-compute process, and then modify the nova/cmd/__init__.py file "Eventlet.monkey_patch (Os=false) "for" Eventlet.monkey_patch (all=false, socket=true, Select=true), and finally launch Nova in eclipse The nova-compute process of the Pydev project. Running Devstack is to quickly build a debug environment to focus on the code you want to debug, but regular devstack or other unrelated components like to drag your hind legs. To summarize:

1, perform the first ' Nova Boot ' command can start a virtual machine, but the second time to execute directly in the Nova-schedule there is the error, the machine configuration is OK, the resources must be enough. So debug some, the problem found, in the default section of the nova.conf file to add the configuration "service_down_time = 7200", done. That is because the debug time is long, the Nova-compute process does not report to the DB in time it is still alive, so nova-schedule mistakenly think that there is no suitable compute node to dispatch.

2, the direct operation in Eclipse is not a problem, but as long as a debug to call Nova-conductor method when the breakpoint will hang there, and occasionally appear in the log "messagingtimeout:timed out waiting for a Reply to the message ", to point to the absolute, directly modify the nova.conf file, add:

[Conductor]
Use_local=true

If the compute node goes down, but does not disable the host in Nova, Service_down_time and Report_interval Setting time Nova-schedule will mistakenly think that the host is still alive, and thus a problem. In addition, it could be a olso bug, https://bugs.launchpad.net/oslo.messaging/+bug/1338732

Or remove the Retryfilter,

Scheduler_default_filters=availabilityzonefilter,ramfilter,computefilter,computecapabilitiesfilter, Imagepropertiesfilter

3, I downgrade the code from the master branch to Icehouse after I run devstack when I migrate DB's pin this is not to say that this is missing that, delete the high version of the PYc file, done.

Find. -name "*.pyc"-exec rm-rf {} \;

4, in my impression, run devstack, Glance is the most like the components of the convulsions, no time to talk about it, that in the stack.sh directly commented out upload_image that line, and finally manually add the mirror.

wget http://cdn.download.cirros-cloud.net/0.3.2/cirros-0.3.2-x86_64-disk.img
Glance image-create--name=cirros-64--disk-format=qcow2--container-format=bare--is-public=true--progress < Cirros-0.3.2-x86_64-disk.img

5, and Os-queues-api-version related error, that is because now Marnoi renamed Zaqar, need zaqarclient instead of marconiclient, so: sudo pip uninstall python_ Marconiclient . Or python-openstackclient will scan/usr/local/lib/python2.7/. All openstack.cli.extension modules in the Dist-packages directory go to try to load, so zaqarclient and marconiclient two are loaded repeatedly.

Plugin_modules.extend (Get_plugin_modules (
' Openstack.cli.extension ',
))

The result is that the Keystone Create_keystone_accounts script will be reported as "Argparse. Argumenterror:argument--os-queues-api-version:conflicting option String (s):--os-queues-api-version " The result is that there is no initial data in the database, and eventually you see the error:

/bak/openstack/devstack/functions-common:286:die
2014-12-23 01:38:17.229 | [ERROR]/bak/openstack/devstack/functions-common:1192 Keystone fail to get token

6, Openstack-client and other library versions, such as Python-neutronclient, can be configured by configuring the Libs_from_git configuration to let the python-neutronclient of these libraries go git to install the library to/ Usr/local/lib/python2.7/dist-packages/python_neutronclient-2.3.10.post2-py2.7.egg-info
Running install_scripts, not pypi.

Libs_from_git=python-neutronclient,neutron-vpnaas,neutron-fwaas,neutron-lbaas,python-keystoneclient, Python-glanceclient,python-novaclient,python-cinderclient

Will sometimes use the openstack-client command times to make the following error:

+ + + OpenStack Project Create admin--or-show-f value-c ID
2014-12-23 01:44:30.542 | ERROR:openstackclient.shell Exception raised:python-neutronclient 2.3.9.40.g9ed73c0 is installed but python-neutronclient<3,>=2.3.6 is required by []

7, the modules in Oslo are also prone to problems , making sure to use the latest code.

For Olso_pro in ' Pip Freeze |grep oslo |awk-f ' = = ' {print $} '; Do
echo ' Upgrade ' $olso _pro
sudo pip uninstall-y $olso _pro
sudo pip install--upgrade $olso _pro
Done

8, sometimes, sudo python setup.py install installs the data file into the/var/lib/etc/neutron directory, and the--prefix=/can be loaded into the/etc/neutron directory.

9, sometimes the OpenStack source code from one machine to the other machine to run, this line error one of the following SSL error, need: RM-RF. /requirements/.venv/

Env http_proxy= https_proxy= no_proxy= pip_find_links=file:///bak/openstack/.wheelhouse/bak/openstack/requirements /.venv/bin/pip install-u PBR

10, the virtual machine cannot ping the host. Check whether the eth0 of the virtual machine is added to the OvS network in a non-internal way, and the route is configured on the eth0

sudo ovs-vsctl----may-exist add-port br-phy eth0--set interface eth0 type=internal

http://blog.csdn.net/quqi99/article/details/40949799

Major issues encountered when debugging OpenStack (by quqi99)