I am grabbing nodes from MaaS and and deploying my Juju bundle, but cannot spawn containers, because whether I tell it to or not, it wants "Network Spaces" data which I cannot turn off, and am thus far unable to provide it. Here is my status output:
Here is my bundle:
Ubuntu 16.04.3 LTS
During the latest batch of updates I was informed that the LXD PPA packages were being removed in favor of just installing as a snap.
│ LXD PPAs to go away by end of year │ │ │ │ We are deprecating all LXD PPAs at the end of 2017. │ │ │ │ Existing users should move to the LXD snap as the preferred way to get the latest LXD │ │ feature release on older Ubuntu releases. │ │ │ │ You can do so by first installing snapd on your system if it's not there already. Once │ │ snapd is installed, installing the LXD snap and migrating your existing data can be done │ │ with: │ │ │ │ snap install lxd && lxd.migrate
I am running bridged networking with some ugly routing rules, a 12TB ZFS RAIDz pool that takes about 6 weeks to restore from the cloud using Crashplan, and some other fun stuff to allow certain containers access to local devices and mount points. I'm scared to migrate because even running through a LXD conjure-up was tricky on a clean install, the snap symlinks do not get set up by default so there is an extra layer of complexity in calling the commands from the snap bin/ directory instead of them being global, and I have a large amount of data data I want to save in many of these containers, etc..
Previously I had an issue where I was forced to migrate the pool and had to update the lxd.db with the new pool name before like so:
echo "UPDATE config SET value='lxd/angel' WHERE key='storage.zfs_pool_name';" | sqlite3 /var/lib/lxd/lxd.db
I can create a new location in the existing pool with one of these:
zfs create lxd/snap
...but isn't there a way to just import the existing config to use what I already have without having to copy all this data back to the same array?
root@ayana-angel:~# zfs list | grep snap lxd/angel/snapshots 384K 3.69T 128K none lxd/angel/snapshots/mysql-server-0 128K 3.69T 128K none lxd/angel/snapshots/plex-0 128K 3.69T 128K none lxd/snap 128K 3.69T 128K none
Also, this doesn't even address the reuse of the existing network bridge. I seem to recall the configs being no fun at all to muck around in, so I really want to keep them.
root@ayana-angel:~# history | grep lxc |grep br0 1049 lxc network attach-profile br0 default eth0 1050 lxc network detach-profile lxdbr0 default eth0 1212 lxc network set br0 ipv4.nat true 2045 history | grep lxc |grep br0
Does anyone have a better way to do this? The last time I jacked this up it took about a week to figure out how to make new container images from tarballs and import them to the new location.
Thanks.
I have a host (Openstack Newton data center controller) with the following services running on it:
These are running as LXD containers on ZFS. The hardware is an Intel NUC 5i5MYHE blade with an SSD drive. The blade is fed a set of VLAN trunks with the first VLAN untagged. This untagged LVAN is our core data center network. I have a Maas-Region-Controller upstream that communicates to the rack controller and manages it.
MaaS has a DHCP reservation excluding the first 10 IP addresses on this network for these core services, and is set to assign addresses from a DHCP pool to LXD Openstack components on my other blades over the untagged VLAN when they PXE boot to the internal rack controller.
I can easily set the LXD bridge interface IP address to DHCP since I technically do not care about it, but the controller host that all of this lives on needs a static address in the reserved range.
Juju will need a DHCP assignment to be able to complete it's bootstrap process.
I am searching for something like this, although this doesn't work for obvious reasons:
# The loopback network interface
auto lo
iface lo inet loopback
# The primary network interface
iface enp0s25 inet manual
address 192.168.199.5
netmask 255.255.255.0
network 192.168.199.0
broadcast 192.168.199.255
gateway 192.168.199.1
# dns-* options are implemented by the resolvconf package, if installed
dns-nameservers 192.168.199.1 192.168.1.1 8.8.8.8
dns-search home.spyderdyne.net
auto br0
iface br0 inet dhcp
bridge-ifaces enp0s25
bridge-ports enp0s25
up ifconfig enp0s25 up
Any thoughts on a way to allow DHCP across the bridge with a blade having a static IP and not breaking VLANs are welcome. I am thinking about some sort of sub-interface definition maybe.
Obviously this is not covered anywhere that I can find and probably represents an odd setup, but I am trying to make the most of limited resources and space for this project.
I have a small lab (literally) of 3 RBP-3B units with SATA and SD drives attached running Ubuntu 16.04 LTS. I have installed software on them to create a Ceph cluster but its tricky and pretty fragile.
Ultimately I would like to administer this "Worlds Smallest Cloud Storage Cluster" concept with Juju and make it the storage backend for an Intel NUC vBlock I won from the OSIC contest at the Openstack summit. Unfortunately I cant find a way to PXE boot the Raspberry Pi 3B, and chip support for this feature is on the horizon but currently not present.
I want to use MaaS to bootstrap the OS and feed these little guys into Juju deployments since its so easy from there to manage everything, but I am having trouble finding a working solution. I know that Juju can add machines after the OS is installed but its not apparent how to do that either, and if I set Juju up to use existing machines I don't know if I can still point it to MaaS controlled systems as well to create a heterogenous mix of gear without breaking functionality.
Any thoughts?
I have multiple Ubuntu Linux hosts that need to connect to Active Directory for user management (unfortunately an actual LDAP server was not an option here...) and have configured Winbind to handle authentication, and my %sudoers group entry works fine:
## Allow AD members of the service group restricted user access to that account
%sudoers ALL=(ALL:ALL) NOPASSWD:ALL
I have allowed password authentication in the ssh config and have added the following entry to my sudoers file for the account:
%service ALL=(service:service) NOPASSWD:ALL
This doesn't even find the group even though the sudoers group and the service groups are at the same level in the directory tree on the AD server. When I do this:
%domain.local\\service ALL=(ALL:ALL) NOPASSWD:ALL
...members of the service group are able to log in and have full sudo rights, including root which is unacceptable, and already covered by the sudoers group and entry. Alternatively when I do this:
%domain.local\\service ALL=(service:service) NOPASSWD:ALL
...members are unable to sudo to any user. As for the service user, here is the /etc/passwd entry for the service user:
service:x:1001:16777230::/opt/service:/bin/bash
Per the Sudoers Man Page this last entry should probably work, but for some reason it doesn't. Also, the %service should also probably cover both local and AD provided membership but it doesn't seem to. When I create this user since there is already a group in the directory server named service here is how I am creating the service user account:
useradd -d /opt/service -g service service
Here is the output of sudo -l for a member of the group:
[user.test@server ~]$ sudo -l
Matching Defaults entries for user.test on this host:
requiretty, !visiblepw, always_set_home, env_reset, env_keep="COLORS
DISPLAY HOSTNAME HISTSIZE INPUTRC KDEDIR LS_COLORS", env_keep+="MAIL PS1
PS2 QTDIR USERNAME LANG LC_ADDRESS LC_CTYPE", env_keep+="LC_COLLATE
LC_IDENTIFICATION LC_MEASUREMENT LC_MESSAGES", env_keep+="LC_MONETARY
LC_NAME LC_NUMERIC LC_PAPER LC_TELEPHONE", env_keep+="LC_TIME LC_ALL
LANGUAGE LINGUAS _XKB_CHARSET XAUTHORITY",
secure_path=/sbin\:/bin\:/usr/sbin\:/usr/bin
User user.test may run the following commands on this host:
(service : service) NOPASSWD: /opt/service
What am i missing here?
I am running Ubuntu 16.04 and am up to date. When I follow the instructions here for a single machine installation:
Openstack Single Installer Guide
Here is the sequence:
apt-add-repository ppa:cloud-installer/stableapt-get updateapt-get install -y openstackIncredibly (possibly part of the problem...) after only 308MB of package and component downloads this completes successfully. Unfortunately the next step fails:
ubuntu@xenial-64-base:~$ openstack-install
The program 'openstack-install' is currently not installed. You can install it by typing:
sudo apt install openstack
The suggestion to apt install openstack shows that it is already installed, and the latest version. I can't tell for sure if this is a problem with the repo, the installer, or with the current Xenial updates having issues with something older in the ppa. I am left with an openstack executable which takes me to a prompt but the openstack-install exec seems to be missing:
ubuntu@xenial-64-base:~$ openstack
(openstack) install
This is the definition of optimism...
openstack: 'install' is not an openstack command. See 'openstack --help'.
Did you mean one of these?
catalog list
catalog show
console log show
console url show
(openstack) catalog list
Missing parameter(s):
Set a username with --os-username, OS_USERNAME, or auth.username
Set an authentication URL, with --os-auth-url, OS_AUTH_URL or auth.auth_url
Set a scope, such as a project or domain, set a project scope with --os-project-name, OS_PROJECT_NAME or auth.project_name, set a domain scope with --os-domain-name, OS_DOMAIN_NAME or auth.domain_name
Since it is complaining about not having an RC file we have the python clients installed but without a working installer nothing is configured.
According to Launchpad there have not been any commits since the release 7 weeks ago:
I can try manually performing the steps that openstack-install if I can figure out what they are. Openstack Autopilot is not an option since I don't have the hardware prerequisites sitting around, and ultimately the intent of this is to provide a single image for integrating Openstack test suites (Rally, Tempest, Browbeat, Mythos) for deployment to multiple Openstack deployments around the globe and a tiny footprint is required to make it attractive.
Is anyone else aware of this issue and has anyone found a workaround?
I am writing a script that downloads some files and executes them. I only want each filename executed once by the hosts that are downloading them. I am trying to create a script that pulls the files, executes them if they havent already been executed, and then keeps track of each file that has been executed to prevent duplicate runs. So far all I seem to be able to do is execute all each time, and the backup record does not trigger the deletion of the duplicates.
(cd /opt/trunk/mythos/medusa/remote-scripts/ && wget -r --no-parent --reject "index.html*" http://server/medusa/scripts/)
chmod +x /opt/trunk/mythos/medusa/remote-scripts/server/medusa/scripts/*.sh
for each in /opt/trunk/mythos/medusa/remote-scripts/server/medusa/scripts/*.sh
do
if /usr/bin/test -e $each.bak
then
rm -rf /opt/trunk/mythos/medusa/remote-scripts/server/medusa/scripts/$each
fi
bash $each
mv $each $each.bak
done
What am I doing wrong?
I need to generate vagrant images that are configured as master/slave types. The slaves need to have SSH access to the master node, and there will be tens of thousands of slave instances.
Vagrant supplies Puppet and Chef provisioners, so I used Puppet since it is already floating around in the infrastructure so it should be easier to maintain here.
I am performing the following to generate SSH keys on the node that spawns the instances that will become my images like so:
ssh-keygen -t rsa -N "" -f keys/id_rsa_medusa
Unfortunately this seems to be generating SSH keys that still require a password. I am also generating these keys on an Ubuntu host and using puppet to pull them in to the vagrant instances, which does not seem to work.
What is a 1 liner or script I can use that will actually generate an ssh key with an empty passphrase?
Note:
This doesnt work,
echo -e '\n \n' | ssh-keygen -t rsa -f keys/id_rsa_medusa
BONUS:
ssh-keygen is creating keys with the base host's identity. Generally you would generate these from the host you are trying to ssh from and ssh-copy-id them over to the host you are trying to connect to. In this scenario, the host ID of the machine generating the keys is neither host. How do you generate generic keys that allow any host to access the server host?
Since this is vagrant (headless puppet) I am unable to copy keys between instances because somehow they both have the same IP address when I log into them. I have attempted to generate keys on the base host, then use the following to convert to PEM format, but it doesnt work either:
ssh-keygen -f keys/id_rsa_medusa -e -m pem > keys/id_rsa_medusa.pem
Thanks.
We are attempting to deploy Ubuntu Cloud Openstack Juno with MAAS and need to determine each Internet endpoint that Ubuntu Cloud needs to connect to from the LAN. Each destination needs to be white listed and categorized or we are unable to access them.
While the Ubuntu Cloud installation instructions provide a list or additional Apt repositories, there is no mention of what Juju needs access to in the Internet, or anything else that is not in the repository list.
Netstat outputs for the locations I have managed to catch already:
root@hostname:~# netstat -untap | grep 443
tcp 0 1 192.168.0.x:49128 91.189.89.222:443 SYN_SENT 57117/python3
root@hostname:~# netstat -untap | grep 443
tcp 0 1 192.168.0.x:44643 91.189.89.223:443 SYN_SENT 57026/python3
How can I find the rest?
root@spsandboxbl02:/var/log/maas# cat /etc/issue
Ubuntu 14.04.1 LTS \n \l
Following this guide:
http://maas.ubuntu.com/docs/install.html
root@spsandboxbl02:/var/log/maas# apt-get install maas maas-cluster-controller maas-dhcp maas-dns
Reading package lists... Done
Building dependency tree
Reading state information... Done
maas is already the newest version.
maas-cluster-controller is already the newest version.
maas-cluster-controller set to manually installed.
maas-dhcp is already the newest version.
maas-dns is already the newest version.
0 upgraded, 0 newly installed, 0 to remove and 10 not upgraded.
I have a MAAS server up and running with 3 networks;
MGMT - x.x.x.0/26 DNS and DHCP Managed PUBLIC - y.y.y.0/26 Unmanaged DATA - z.z.z.0/26 Unmanaged
I have the MAAS Cluster manager node up and running, PXE boot images downloaded, but am unable to join nodes to it. My additional nodes are using eth0 and should be receiving DHCP from MAAS but are not. Even manually restarting the dhcp server in the MAAS node has no effect. As a result PXE is also not usable.
Symptoms:
Manually adding a node:
no communication
PXE booting a node on the MAAS managed net:
No DHCP address received (apt-get install maas-dhcp maas-dns completed, dpkg-reconfigure maas-dhcp and MAAS performed as well)
Booting from the install disk as MAAS install:
No DHCP, manual networking config on the subnet completed with a sigterm, node doesnt come back up or provide a status message of any type, and has nothing on it after 6 hours.
I have found no way to add nodes to MAAS using Ubuntu 14.04 (no vlans) so far.