Age | Commit message (Collapse) | Author | Files | Lines |
|
Bug where default gateway was always being set to the first IP in the
subnet. Now should detect correct public gateway. This only impacts
the default gateway configured on the external neturon
"provider_subnet".
JIRA: APEX-37
Change-Id: I0296fa7149679a0f2eb47f7d65e0011f366c1147
Signed-off-by: Tim Rozet <trozet@redhat.com>
(cherry picked from commit f5032bce67f6b7b8f8c5186b6f64e83bd113338e)
|
|
|
|
Puppet relies on facter to gather host facts, including domain name.
Unfortunately facter gathers domain name by looking at the first domain
listed in /etc/resolv.conf, rather than the domain name set on the host.
This fix corrects the domain in that file.
JIRA: APEX-36
Change-Id: If4e4a0b9f82e562463d97ceeee6b7dfd445e3bdc
Signed-off-by: Tim Rozet <trozet@redhat.com>
(cherry picked from commit d4ccd3bdf95f4250747f72d0ca98a19f3656d128)
|
|
Simply use a new argument "-single_baremetal_nic <nic name>" to define which NIC to
bridge to on your host. Also ensure when using this arg that your
ksgen_settings file has mac_address and private_mac as the same value
for your controllers (since you are using a single interface).
JIRA: APEX-33
Change-Id: I69095ccc4b907ae3c6ea3e88b2d1c9af9894c5cb
Signed-off-by: Tim Rozet <trozet@redhat.com>
(cherry picked from commit f6947d7d35d5b0dd3a8a38152a367d9422b81279)
|
|
BGS-84
After the changes to br-ex on the control nodes. The default route
was no longer correct on the control nodes in bare metal deployments.
Added an admin_network paramater for barmetal deployments.
Added logic on br-ex in to set DEROUTE=no on the admin network
in the extern_net_presetup.
Tested on HA physical deployment.
Change-Id: Ic4b1ef521da2c55590990411f3f2444d4c94b019
Signed-off-by: randyl <r.levensalor@cablelabs.com>
(cherry picked from commit 7dc718a545c2da1b7bf150bf652705aeb98b245c)
|
|
Users may now provide domain_name in the ksgen settings file (along with
setting the correct and respective hostnames). This domain name will be
used for all of the deployed machines.
JIRA: BGS-86
Change-Id: I34b6da07dd565648399ea7a1abf2abe1ae62d91a
Signed-off-by: Tim Rozet <trozet@redhat.com>
(cherry picked from commit 1089c79563e61589d105accc3067f6ebbd3020ed)
|
|
Change-Id: Ifb1ab47017ab28fab0a39961c2fd7728d2098cda
Signed-off-by: jose.lausuch <jose.lausuch@ericsson.com>
(cherry picked from commit cf15cf80302559a1647d243187240cfca1efaad2)
|
|
VM resources (vcpus, memory, disk size) are now configurable in the
opnfv_ksgen_settings files. Each node in the ksgen settings file now
has "memory, cpus, disk" attributes:
- cpus is an integer number of CPUs to assign to the VM
- memory is an integer size in kilobytes
- disk is an integer size in gigabytes
The vagrant box vmdk is resized for each VM. Each VM will then modify
its partition to the new size, execute the normal reboot, then extend
the physical+logical volumes and filesystem using vagrant provisioning
scripts.
JIRA: APEX-1
Change-Id: I471319db14d2189052961b2f533c65850ebeb1e7
Signed-off-by: Tim Rozet <trozet@redhat.com>
(cherry picked from commit 5384a722ffae47bc3519512d34802102ebc7be5b)
|
|
The printed horizon URL at the end of a successful deployment was
incorrect for non-HA deployments. The private IP was being printed
which is not publicly accessible. This patch fixes the URL printed to
be the public IP of the control node - the correct URL.
JIRA: APEX-12
Change-Id: Ia2f594e0cf839a9100c3bf5715a38fe4fd243bfe
Signed-off-by: Tim Rozet <trozet@redhat.com>
(cherry picked from commit db4a9ac5492659cd435b4e4e84868510259c6b29)
|
|
Patch adds a check to make sure that the user specified static_ip_range
argument (Ex. 192.168.0.10,192.168.0.20) falls within the valid public
subnet of the host (the same subnet as the default gateway).
JIRA: APEX-19
Change-Id: Ifc0b79074b887509617c2749e50d97d1800abad0
Signed-off-by: Tim Rozet <trozet@redhat.com>
(cherry picked from commit 50fe670726f2229ef963517a890fef2d821b8821)
|
|
Issue where the wrong field was being used for finding interfaces.
JIRA: APEX-11
Change-Id: I94635837c73acd0d271dbc9d0474dc5ba688ff10
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
|
|
deploy.sh was failing in user deployments because the necessary puppet
modules and vagrant files were being copied accidentally from /tmp and
not the script directory. This patch fixes that issue. clean.sh was
failing checks to see if vagrant or virtualbox was still running, due to
a bug with checking "ps" and using grep. This patch resolves that.
JIRA: APEX-16,APEX-17
Change-Id: I1faa6fc134c0308acb2e7b14be30f7cd3c99d109
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
Removes default gateway route to 10.0.2.2 after adding public default route
JIRA: APEX-2
Change-Id: I7ba23e916f3995f39ee1677007ae5c1612cb3cc9
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
|
|
This patch deprecates the use of puppet-trystack repo and migrates the
puppet modules even to Genesis. deploy.sh now now copies the modules
from the current Genesis repo into foreman VM's /vagrant to be used
during install (rather than recloning). Associated changes in
Khaleesi/QuickStack:
- Khaleesi: commit 31b912778847f295d1459a71e5c41c808977c91d
- Astapor: commit ac4be11e348d7bcff1e1aa28c96056e18dba5291
JIRA: APEX-8
Change-Id: I322ac797d1d687f17434e07e6775acef8961bdef
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
This patch addresses the detection of the physical NICs to bridge to.
There was a bug where the regexp would ignore some NICs on accident, and
the purpose of the expression was to ignore virtual NICs. Rather than
ignore NICs by their name, this patch ensures that NICs are ignored if
they are actually virtual types.
Also fixes syntax issue with checking for nic_arg_flag.
JIRA: APEX-11
Change-Id: Ibd965c454faab1d751e3ce0b260eabc7fbe7c720
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
|
|
|
|
|
|
Chef removed the centos7 basebox from Atlas. A new opnfv base box was
added. This patch includes the modifications necessary to point to
that new box.
JIRA: APEX-14
Change-Id: I7e74726c692f21583d0d70f9dd7558665dfb5d99
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
Ensures nodes are shutoff if a baremetal deployment, before attempting
to deploy. Previous behavior is deploy will run up to Foreman pxeboot,
then fail. We can save a lot of time by detecting this up front. Nodes
must be shutoff before deploying.
JIRA: BGS-68
Change-Id: Ie585e12a5271bc1baac60f6256f85e5cb1b1baeb
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
This patch adds new functionality to be able to specify the physical
interfaces on the jumphost to bridge to (rather than figured out
dynamically). The patch introduces new args for each nic:
- admin_nic
- private_nic
- public_nic
- storage_nic
The public_nic is the only arg to be used when using -virtual. In
baremetal: admin_nic, private_nic, and public_nic must either be all or
none specified. The patch also cleans up virtual node logic, which was
working, but had unnecessary logic.
JIRA: APEX-4
Change-Id: Ic0a2e1d8ce2b7464441e3e9041faf123b86ca7e6
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
New check makes sure that 3 controller nodes are defined in base_config
when ha_flag is set to true in base_config.
JIRA: APEX-7
Change-Id: I7f49dec82704d8c9cbbcf17eb004b0adede406b3
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
Fixes clean so that it does not hang when base_config is not provided.
Also now removes libvirt libraries which conflict with VirtualBox
required for deploy.
JIRA: BGS-78
Change-Id: I7c71ca207f260eafb1fb97ace764b9d288fc258a
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
JIRA: OCTO-109
Change-Id: I272a1f9f12454ff71a8a1915ab7b712474b84832
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
|
|
Verifies that subnets are at least these sizes for deploy:
- admin: 5 IPs
- private: 10 IPs
- public: 25 IPs
- storage: 10 IPs
JIRA: BGS-71
Change-Id: I0a6c373c9a8737fa9f0c2712851616dc301968f3
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
Now all VMs in the vm_dir (/opt/var/opnfv) will be destroyed and
removed. Also, virtual box VMs that are leftover even if Vagrant is
stuck will attempted to be gracefully removed first, before forcefully
destroying the process. The clean should work for both baremetal and
virtual deployments.
JIRA: APEX-3
Change-Id: Ib31cf1210651e4413ec8fe918ff91be63f4ad6d4
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
For bare metal deployments, the list IP of the last IP of the subnet
and 20 IPs before are reservered for neutron public IPs.
A new function to subtract a number from an IP was added.
Really fixed the logic for skipping 10 IP for VIPs was flipped to treat
no_dhcp with static IPs as a special case.
Removed a false print statement that the foreman VM is running.
Consolidated two identical print messages with the public IP range.
Change-Id: I189b9440dafd98e49d9a147515810e44b97ab256
Signed-off-by: randyl <r.levensalor@cablelabs.com>
|
|
Looks like for some reason GIT_SSL_NO_VERIFY=true is required for the
git clone to work in LF lab. I guess git cannot recognize the SSL
cert. However, this same git clone command works on another non-LF
server that is using the same exact git version without the need for
this fix.
JIRA: BGS-60
Change-Id: I2a71902b3fe01026644ada29bef67c3db3cda711
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
JIRA: BGS-60
Change-Id: I23a004d41ff8fd37a0ddcae1b4831a333000cae8
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
JIRA: BGS-60
Change-Id: Ida9458cf638eca81e6a3d67941203c605d9e4a8a
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
No longer relies on bgs_vagrant github project. Now Foreman VM is
created under /var/opt/opnfv/foreman_vm, instead of /tmp/bgs_vagrant for better
naming convention and file location. Other VM nodes are also placed in
/var/opt/opnfv/.
JIRA: BGS-60
Change-Id: I009d907ad777750168b822ab86cd11515e28cdd7
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
|
|
This patch will take care of the subnet on public range. Instead of
having a stoned /24, we can use now use at least /27 subnet (because
foreman will try to use 20 IPs for public floating ip pool)
This is not the best way to do that, but it's better than the current
way.
It also add a parameter to set the number of floating IP we want to use
from the public subnet in provider network.
Change-Id: I467f2a4098d2da3c6f666453cead64e18d0c655c
JIRA: BGS-75
Signed-off-by: arnaudmorin <arnaud.morin@gmail.com>
|
|
No longer needed post Arno
Change-Id: I24ba94e48658372034e334d413f044f407f2cb72
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
For the LF pods, it seems that the ansible reload playbook is necessary,
in our pod, we also need to wait 1 minute before trying to run shell
provisionner in Vagrant VM.
This patch adds a 1 minute pause into the ansible playground in order to
be sure that the machine will be accessible.
This avoid "SSH connection unexpectedly closed" issue.
Change-Id: Iab2b47decb0120e8359a175b1f16ccbd1036e91f
JIRA: BGS-85
Signed-off-by: arnaudmorin <arnaud.morin@gmail.com>
|
|
Patchset fixes the following:
- Virtual deployment no longer requires 3 networks, does a lookup of
the public interface on the server to bridge to, and then uses internal
VirtualBox networks for the other admin and private networks
- enable_virtual_dhcp option now allows enabling of Foreman DHCP
server. This new default behavior is to use static IP address in
virtual deployments. This is more useful in environments where
disabling DHCP on the public network is not an option.
- static_ip_range argument used to define which IP block to use out of
the public network for the deployment. Valid only in virtual deployment.
- puppet verification for virtual deployments added
- horizon and foreman URLs will now be printed at the end of deployment
- fixes and updates to non-HA deployments
- ping_site argument now allows users to specify the site to verify VM
internet connectivity. Default is google.com which is unreachable in
China
JIRA: BGS-73
Change-Id: I63b4c91477591d2f9436b5e6f59a2f2ee021d7d7
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
To help with readability and troubleshooting, the 80% of the
script that was not in a function were added to functions
based on logical blocks.
Neither functional nor order of execution changes were made
to the script.
Since most of the file and other Foreman CI scripts use a
2 space indent, that was also applied uniformly across the
file.
Change-Id: I83b96e231d484813025f6aa900cf2ffc92e94397
Signed-off-by: randyl <r.levensalor@cablelabs.com>
|
|
With the foreman install the vboxdrv module is not loaded
then the deploy.sh script attempts to load the driver by
running ‘/etc/init.d/vboxdrv setup’
and checks the return code for success.
However, /etc/init.d/vboxdrv will return 0 as long a valid parameter was passed on the command line. In the case of a
failure, a failure message will be logged and the return code
will still be 0.
For instance if the kvm module is already loaded, the
VirtualBox driver will never install.
deploy.sh will now check for the VirtualBox kernel module
with lsmod after the setup script is installed.
The deploy.sh will now exit if the vboxdrv is not loaded.
Change-Id: I702819cbf28afb08e0035e08918390af85c07674
Signed-off-by: randyl <r.levensalor@cablelabs.com>
|
|
tools
JIRA: BGS-56
Change-Id: I4af161c7294c7a576920a8400b2d608ccbcbb323
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
Software released under 'Arno' should be pinned to at least
the major version, to guarantee API backwards compatibility.
I believe this commit covers the remaining dependency installs
done by Foreman.
JIRA: BGS-54
Change-Id: I616f4d201dab2ebf6c0b38b9a5aaa1510ba59bfd
Signed-off-by: Daniel Farrell <dfarrell@redhat.com>
|
|
Software released under 'Arno' should be pinned to at least
the major version, to guarantee API backwards compatibility.
This also moves some pip-based installs to yum, for simplicity
and consistency.
JIRA: BGS-54
Change-Id: If4f20a5740b989bd7e369a2ec5aefd559eb505b1
Signed-off-by: Daniel Farrell <dfarrell@redhat.com>
|
|
Software released under 'Arno' should be pinned to at least
the major version, to guarantee API backwards compatibility.
JIRA: BGS-54
Change-Id: Iad027f1dc40b17c611ef183ffbae834b0dd1b970
Signed-off-by: Daniel Farrell <dfarrell@redhat.com>
|
|
This code was developed in a scratch space GitHub
repo, mostly by Tim. As part of the clean-up process
for Arno, it should be migrated to Genesis and all
future work should be done via Genesis.
This is trozet/bgs_vagrant as of f27548.
I didn't copy the clean.sh, deploy.sh and build.sh scripts
from bgs_vagrant in this commit. They differ from those
in Genesis and need more attention for a proper migration.
See: https://github.com/trozet/bgs_vagrant
JIRA: BGS-53
Change-Id: I512e0ea0d02f8d99048db771221abc88aa60e2d5
Signed-off-by: Daniel Farrell <dfarrell@redhat.com>
|
|
|
|
Deploy will now detect and set extra parameters in the ksgen settings
file needed in order to configure an external public network for
Neutron. The new parameters required are as follows:
- external_network_flag: "true"
- public_gateway:
- public_dns:
- public_network:
- public_subnet:
- public_allocation_start:
- public_allocation_end:
- deployment_type:
public_* parameters are required to create an external network when
external_network_flag is "true". deployment_type (three_network or
multi_network) is now passed to Foreman. Foreman decides on this value
whether to configure 3 interfaces as dhcp+onboot or 4.
Change-Id: Ia8af04e82f3854a36adc563c1855d451ddfa9c17
JIRA: BGS-13
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
This correct a small typo that prevent the clean script to kill
VBoxHeadless correctly
This also add the psmisc package to dependency list in order to have the
killall command
JIRA: BGS-49
JIRA: BGS-50
Change-Id: I39143bdbc8076318c5a28868c4bb779411742c09
Signed-off-by: arnaudmorin <arnaud.morin@gmail.com>
|
|
This patch adds the argument "-virtual" which allows one to deploy to
Vagrant VMs rather than baremetal. The requirements for the Jumphost
become slightly different with VM nodes. The required RAM increases to
24GB when using virtual. Usage: "./deploy.sh -virtual".
JIRA: BGS-31
Change-Id: I2e5e60cc7913f4d3efaa405950e1fcbbefd2771b
Signed-off-by: Tim Rozet <trozet@redhat.com>
|