Aiab multinodes: Replacing virsh by Openstack VMs
Hello Airshipers, Still experimenting with Airship, after a successful Singlenode deployment, I'm trying the multinodes Site. My aim is to be able to use the Airship built on top of Openstack VMs and not on real Hw nor AIO virsh VMs on a single node... So that I could increase VM size to install other Sw at the upper layer and troubleshoot easily what's going on below. So, I remove the Virsh network and VMs creation from the first stages (create_VMs, gate_setup) of the gate/multinode_deploy and replace them by Openstack VMs of the expected Size and with expected Network connexion and Fixed addresses 172.24.1.xx on an Openstack Network. With this config, except a dns conflict fixed, it works and I get a Genesis/Airflow access. (Yeah!) As you can figure out, things became more Airship specific later on, because in order to setup the nodes, the Workflow task "prepare_and_deploy_nodes" try to make them boot for pxe, using a libvirt driver. I've found in the deployment_files.yaml that configuration which seems to be the blocking point: data: hardware_profile: 'GenericVM' primary_network: 'gp' oob: type: 'libvirt' libvirt_uri: 'qemu+ssh://virtmgr@172.24.1.1/system' So I think I have following options and I would prefer the simple one : - either replace the libvirt drydock driver in the site definition by another one (redfish?): But can the redfish driver drive Openstack VMs (In this case I need to add a Redfish Server VM ?) - or, remove the prepare_and_deploy_nodes task : But is it simple or need to modify Drydock code? Is it really feasible? Any simpler option is welcome... Thanks in advance, Loïc
Hello Loic, No, RedFish driver would not be able to deploy OpenStack VM. RedFish is just an alternative to the IPMI, it can change power state of node and set next boot to PXE. May be try manual driver? But by default it waits for 1 minute only, might be not enough. The prepare_and_deploy_nodes workflow task is defined in Shipyard DAG src/bin/shipyard_airflow/shipyard_airflow/dags/drydock_deploy_site.py. Best regards, -- Roman Gorshunov On Thu, May 23, 2019 at 5:09 PM Loic Le Gal <l.legal.astellia@gmail.com> wrote:
Hello Airshipers,
Still experimenting with Airship, after a successful Singlenode deployment, I'm trying the multinodes Site. My aim is to be able to use the Airship built on top of Openstack VMs and not on real Hw nor AIO virsh VMs on a single node... So that I could increase VM size to install other Sw at the upper layer and troubleshoot easily what's going on below.
So, I remove the Virsh network and VMs creation from the first stages (create_VMs, gate_setup) of the gate/multinode_deploy and replace them by Openstack VMs of the expected Size and with expected Network connexion and Fixed addresses 172.24.1.xx on an Openstack Network.
With this config, except a dns conflict fixed, it works and I get a Genesis/Airflow access. (Yeah!)
As you can figure out, things became more Airship specific later on, because in order to setup the nodes, the Workflow task "prepare_and_deploy_nodes" try to make them boot for pxe, using a libvirt driver.
I've found in the deployment_files.yaml that configuration which seems to be the blocking point: data: hardware_profile: 'GenericVM' primary_network: 'gp' oob: type: 'libvirt' libvirt_uri: 'qemu+ssh://virtmgr@172.24.1.1/system'
So I think I have following options and I would prefer the simple one : - either replace the libvirt drydock driver in the site definition by another one (redfish?): But can the redfish driver drive Openstack VMs (In this case I need to add a Redfish Server VM ?) - or, remove the prepare_and_deploy_nodes task : But is it simple or need to modify Drydock code?
Is it really feasible? Any simpler option is welcome...
Thanks in advance,
Loïc
_______________________________________________ Airship-discuss mailing list Airship-discuss@lists.airshipit.org http://lists.airshipit.org/cgi-bin/mailman/listinfo/airship-discuss
Hey Loic, Great to hear that you're getting your environment working!! The only simpler option I can think of to solve for running in OpenStack VMs might be: 1. deploy a vanilla Kubernetes cluster on your VMs (e.g. standard KubeADM) 2. run the Airskiff scripts (skipping the initial one single-node k8s install) Basically, Airship with a "bring your own infra and kubernetes". This skips the Airship bare metal provisioning (like you want) with the downside of skipping over Airship k8s provisioning. However, it should work well for testing declarative workload (e.g. OpenStack) deployment and lifecycle management inside of your VMs. HTH, let us know how it goes! Matt -----Original Message----- From: Roman Gorshunov <paye600@gmail.com> Sent: Thursday, May 23, 2019 1:02 PM To: Loic Le Gal <l.legal.astellia@gmail.com> Cc: airship-discuss@lists.airshipit.org Subject: Re: [Airship-discuss] Aiab multinodes: Replacing virsh by Openstack VMs Hello Loic, No, RedFish driver would not be able to deploy OpenStack VM. RedFish is just an alternative to the IPMI, it can change power state of node and set next boot to PXE. May be try manual driver? But by default it waits for 1 minute only, might be not enough. The prepare_and_deploy_nodes workflow task is defined in Shipyard DAG src/bin/shipyard_airflow/shipyard_airflow/dags/drydock_deploy_site.py. Best regards, -- Roman Gorshunov On Thu, May 23, 2019 at 5:09 PM Loic Le Gal <l.legal.astellia@gmail.com> wrote:
Hello Airshipers,
Still experimenting with Airship, after a successful Singlenode deployment, I'm trying the multinodes Site. My aim is to be able to use the Airship built on top of Openstack VMs and not on real Hw nor AIO virsh VMs on a single node... So that I could increase VM size to install other Sw at the upper layer and troubleshoot easily what's going on below.
So, I remove the Virsh network and VMs creation from the first stages (create_VMs, gate_setup) of the gate/multinode_deploy and replace them by Openstack VMs of the expected Size and with expected Network connexion and Fixed addresses 172.24.1.xx on an Openstack Network.
With this config, except a dns conflict fixed, it works and I get a Genesis/Airflow access. (Yeah!)
As you can figure out, things became more Airship specific later on, because in order to setup the nodes, the Workflow task "prepare_and_deploy_nodes" try to make them boot for pxe, using a libvirt driver.
I've found in the deployment_files.yaml that configuration which seems to be the blocking point: data: hardware_profile: 'GenericVM' primary_network: 'gp' oob: type: 'libvirt' libvirt_uri: 'qemu+ssh://virtmgr@172.24.1.1/system'
So I think I have following options and I would prefer the simple one : - either replace the libvirt drydock driver in the site definition by another one (redfish?): But can the redfish driver drive Openstack VMs (In this case I need to add a Redfish Server VM ?) - or, remove the prepare_and_deploy_nodes task : But is it simple or need to modify Drydock code?
Is it really feasible? Any simpler option is welcome...
Thanks in advance,
Loïc
_______________________________________________ Airship-discuss mailing list Airship-discuss@lists.airshipit.org https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.airshipit.or g_cgi-2Dbin_mailman_listinfo_airship-2Ddiscuss&d=DwIGaQ&c=LFYZ-o9_HUMe MTSQicvjIg&r=_C5hC_103uW491yNPPpNmA&m=W2xfhYxi7B2i6JWY2my1HMZcj-oFQ0i8 --ik1apG9m8&s=5eA1O44C2m9RU0O1Z_MOAB2wSQYckotwHHjxJu4-rDw&e=
_______________________________________________ Airship-discuss mailing list Airship-discuss@lists.airshipit.org https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.airshipit.org_cgi-2Dbin_mailman_listinfo_airship-2Ddiscuss&d=DwIGaQ&c=LFYZ-o9_HUMeMTSQicvjIg&r=_C5hC_103uW491yNPPpNmA&m=W2xfhYxi7B2i6JWY2my1HMZcj-oFQ0i8--ik1apG9m8&s=5eA1O44C2m9RU0O1Z_MOAB2wSQYckotwHHjxJu4-rDw&e=
participants (3)
-
Loic Le Gal
-
MCEUEN, MATT
-
Roman Gorshunov