Devstack has continued to grow in memory requirements and now we cannot
reliably fit in 4GB, with several services being unable to start. Increase
the minimum for DomU to 6GB to leave room for virtual machines
Change-Id: Idbdfa1f36015b6af347d1ce27eb28baa360af5ef
ANL has discontinued their Ubuntu mirror. Update the target
to be the official round-robin mirror which should be more stable long-term
Change-Id: I1271301d96a44888f45664537435e31dd38ff30d
Dom0's bash does not support associative arrays, however we source
`functions` and therefore our scripts fail. This change breaks the
dependency of dom0 tools on domU functions.
Fixes bug: 1379804
Change-Id: I229e2d5f07070a9236ec612d4032c94c4361a9f6
If you're not using an XVA then the built-VM was previously Saucy, which is
not supported by openstack any more.
Change-Id: I1040d9e43d517582e76f2e1df787986b5ffbc42c
* Ensure that Xen setup scripts will continue to function when unset
variables are used in stackrc
* Ensure that the generic functions are sourced in all places that xenrc
(which sources stackrc) is sourced.
Change-Id: I54eba20733c2e149621b74a1387f0bef14fca12e
Devstack has been increasing in complexity, and if we use too
much of our memory it can lead to fragmentation which in turn
causes a DomU error and a failure of random tests.
Change-Id: Ide9cc84625caed4e35a64a47ee1e92a6cd567651
Driven by a race condition in LVM being hit in the tempest
tests. This is a locking race between a file lock and a
semaphore which is fixed in 2.02.96-5.
Change-Id: I8a8c215c90a0602288292ffd06b7694d2db6219e
xenbr0 is correct for most installations, but not all. Notable xenserver-core
may use a differently named device. Since we can auto detect this, remove the
config and do so.
Change-Id: I989f6ddd5ffb526ab350f263ef6fc402c596304a
In XenServer scenarios, an additional domU is created to run OpenStack
services. This change is increasing the memory for that VM to speed up
test runs.
Change-Id: I322f4e4703e506620fa7e7456c4264ee0d050edc
For Neutron we need an additional OVS bridge to be present. The xe
network-create command did not create the bridge immediately. A
workaround was applied to attach a network interface to that network,
forcing xapi to create the underlying OVS bridge. It turned out, that
the xe network-attach command could be used instead, so the workaround
is removed and replaced with the proper solution by this patch.
Fixes bug 1197305
Change-Id: I10cc763531b71238f0bc12a12de985f8f35d7e27
VM installation created a machine with multiple network interfaces. This
is a needless complexity at that point, we only need one interface. This
change removes the complex network interface setup from
install-os-vpx.sh script, so that only one network interface is created,
which is connected to the selected network (management by default). This
change also removes a lot of non-used code from install-os-vpx.sh
Specify the network to be used for ubuntu netinstall with:
UBUNTU_INST_BRIDGE_OR_NET_NAME
defaults to $MGT_BRIDGE_OR_NET_NAME.
This change also includes:
- use precise as default
Related to blueprint xenapi-devstack-cleanup
Change-Id: Ib0a0a316e849f2c2651305ea657c84820cd2bb3b
OpenStack VM had 4 interfaces, from which eth0 was connected to HIMN (Host
Internal Management Network), and not used at all. The reason for not
using this interface according to johngarbutt is:
- VNC console would not work
- migrations would not work
- many other things
This change removes the interface, thus devstack is providing the
same architecture as described in the official documentation, YAGNI
applied on the HIMN:
http://goo.gl/dmrNF
The change also modifies some defaults, so it's easier to get started:
- The Ubuntu VM is network-installed through the management network
- OS VM is getting its address on the management network through DHCP
- Default Public IP addresses aligned to stack.sh defaults
- OS Services are listening on Management Interface
Related to blueprint xenapi-devstack-cleanup
Change-Id: I73c9751ade6f4786b8b682a1994cb87f0a624379
Some kernels have lockup issues while serving cinder volumes from
file-loopback-lvm kind of cinder setup. This patch provides a way to
create an additional drive to the OpenStack VM, and use that to store
cinder volumes, thus eliminating the lockup issue. It will help when
testing XenServer. Now, you can specify devices for stack-volumes
through:
VOLUME_BACKING_DEVICE
In case you are using CINDER_MULTI_LVM_BACKEND, you can use
VOLUME_BACKING_DEVICE2
as well.
Xenserver:
Should you whish to use a 10 gig disk backend for your cinder volumes,
specify:
XEN_XVDB_SIZE_GB=10
VOLUME_BACKING_DEVICE=/dev/xvdb
Citrix is using this approach on its internal CI system to run tests
against OpenStack. This is a workaround for this bug:
https://bugs.launchpad.net/cinder/+bug/1023755
Related to blueprint xenapi-devstack-cleanup
Change-Id: Iee633d2704185bfbf9234882654c47b850fa168a
This patch makes it easier to support the all-in-one installation of
XenServer with Quantum:
- separate agents for domU and dom0 openvswitches. The domU agent's name
is "q-domua", and the dom0 agent is called "q-agt"
- create a physnet1 physical network to connect those openvswitches
Note: This patch depends on the quantum patch:
https://review.openstack.org/15022
For a diagram, see:
http://goo.gl/BuAdg
Change-Id: Iff851438553528b9051edd380ecbb2c0415cbf78
The Xenserver/XCP part of devstack was configuring the hypervisor's
connectivity to the outer world, by adding VLANs, and physical
interfaces to the virtual networks. It added a lot of complexity, and
made it hard to get started with XenServer. This patch removes that
extra complexity, so it is left as an exercise for the user.
Related to blueprint blueprint xenapi-devstack-cleanup
Change-Id: If3367335c3da8621d0afe1f6cae77511fbdbb3e2
Get rid of UBUNTU_INST_REPOSITORY, use UBUNTU_INST_HTTP_HOSTNAME and
UBUNTU_INST_HTTP_DIRECTORY instead. User can also specify
UBUNTU_INST_HTTP_PROXY to utilize a proxy for the OpenStack VM
installation. The answer file will be edited to contain the specified
values. Also get rid of the magic, undocumented MIRROR variable.
This is related to blueprint xenapi-devstack-cleanup
Change-Id: Ic9fc564c4ad0f43e2e536854335ebe14791d0255
Add error on unitialised variable to the scripts. This way some issues
were identified. Also modify README with fresh variables. The patch
includes:
- Add SWIFT_HASH to the README
- Add XENAPI_CONNECTION_URL to the README
- Add VNCSERVER_PROXYCLIENT_ADDRESS to the README
- Introduce UBUNTU_INST_IFACE which is the OpenStack VM interface used
for the netinstall. It defaults to eth3. Previously this parameter
was set by the combination of HOST_IP_IFACE and the undocumented
NETINSTALL_IP
- get rid of NETINSTALL_IP
- xenrc includes CLEAN_TEMPLATES
- xenrc no longer tries to change directory
- remove chrooting from prepare_guest.sh (STAGING_DIR was always / )
- remove DO_TGZ variable from prepare_guest.sh
- use arguments to call prepare_guest.sh, instead of env vars
- Fix backslash escaping in prepare_guest_template.sh
NOTE: networking is about to be addressed in a separate change.
Related to blueprint xenapi-devstack-cleanup
Change-Id: Ie9a75321c7f41cc9a0cc051398d1e6ec2c88adfa
- allow you to configure the xenapi_user (often other than root)
- allow you to disable the guest installer network
- install the plugins in the xcp-xapi location
- use alternate webserver location when adding the preseed file
- skip the centos specific ip forwarding configuration
- make use xcp inventory, if no xensource-inventory is found
- correctly deal with kpartx to mount the VM VDI in manage_vdi
Change-Id: I8d51725fc97f0bcaa27a46f7a7ced13c369c809e
I have ensured:
- template gets re-used on second run
- template includes XenServer tools, and custom user accounts
- take snapshot before first boot, for easy re-run
- make host_ip_iface work with either eth2 or eth3
- make ssh into domU checks looser
- above is all ground work for improved jenkins tests
- added some more comments to make it scripts clearer
Change-Id: I5c45370bf8a1393d669480e196b13f592d29154f
