Page Comparison

configure connection to nectar server

write and pack the user data

write the user data in a .fcc file

Code Block

variant: fcos
version: 1.4.0
passwd:
  # set user and group for operate on NFS
  groups:
    - name: Q4646RW
      gid: 544646
  users:
    - name: core
      groups:
        - Q4646RW
        # core, adm, wheel, sudo, systemd-journal
storage:
    # write content into file
    files:
      # disable auto updates by config
      - path: /etc/zincati/config.d/90-disable-auto-updates.toml
        contents:
          inline: |
            [updates]
            enabled = false
systemd:
    units:
        # disabling zincati service only works in fcos 36-20220522.2.1 upwards
        - name: zincati.service
          enabled: false
        # mount NFS onto VM
        - name: var-mnt-data-Q4646.mount
          contents: |
            [Unit]
            Description = /var/mnt/data/Q4646
            After=network-online.target
            Wants=network-online.target

            [Mount]
            What = 10.255.122.28:/gpfs/general02/pool9200/Q4646/Q4646
            Where = /var/mnt/data/Q4646
            Type=nfs
            Options= rw,hard,nfsvers=3,nosuid,nodev,noresvport
            TimeoutSec=5

            [Install]
            WantedBy=multi-user.target
        - name: var-mnt-data-Q4646.automount
          enabled: true
          contents: |
            [Unit]
            Description=/var/mnt/data/Q4646

            [Automount]
            Where=/var/mnt/data/Q4646
            TimeoutIdleSec=0

            [Install]
            WantedBy=multi-user.target

generate ign user data file from fcc. docker installed required.

Code Block

language	bash

# fcc file location: C:\Users\uqtlan\fcos.fcc
docker run --rm -i -v C:\Users\uqtlan:/var/opt -w /var/opt \
quay.io/coreos/fcct:v0.18.0 --pretty --strict --output fcos.ign fcos.fcc

specify the parameters for the VM
1. name: name of the instance. drone_pdal_translate
2. port: network port suffix for mounting NFS.
  "-test(-prod)" for port NFSPortnfs-port-02-test(NFSPortnfs-port-02-prod)
3. image: image for the instance.
  tern-fedora-core-38 for running podman directly
4. nets: list of network names. ["qld"]
5. security_groups: list of security group names. ["default", "ssh_public"]
6. availability_zone: availability zone for the VM. QRISCloud
7. key: name of the keypair for the vm. ansible
8. flavor: name of the flavor for creating VM. r3.large
9. user_data: read from ign file as string and base64-encode it

create VM in python by openstacksdk (and push the instance id and ip to xcom)

Code Block

language	py

conn = openstack.connect(**auth)

# user_data needs to be base64 encoded
with open(user_data_file, 'r') as file:
    user_data_raw = file.read()
user_data = base64.b64encode(user_data_raw.encode("utf-8")).decode('utf-8')

# check whether security groups exist
security_groups = []
for security_group in security_groups_list:
    sg = conn.network.find_security_group(security_group)
    if sg:
        log.info('adding security group {}:{}'.format(sg.name,sg.id))
        security_groups.append({"name":sg.name})
    else:
        raise Exception('security group {} not found'.format(security_group))

# check whether image exists
im = conn.image.find_image(image)
if im:
    log.info('using image {}:{}'.format(image,im.id))
else:
    raise Exception('image {} not found'.format(image))

# check whether flavor exists
fl = conn.compute.find_flavor(flavor)
if fl:
    log.info('using flavor {}:{}'.format(flavor,fl.id))
else:
    raise Exception('flavor {} not found'.format(flavor))

# find an available port
port = None
for num in range(1,6):
    suf = port_suf
    _port = conn.network.find_port(f"nfs-port-{num:02}{suf}")
    if _port and not _port.device_id:
        port = _port
        print(f'using port {port.name}:{port.id}')
        break
if not port:
    log.warn("no available port for NFS. using direct mounting instead, which might cause error.")

# nics = [{'uuid': some_net.id},{'port': port.id}]
nics = []
if port:
    log.info('appending port {}:{}'.format(port.name,port.id))
    nics.append({'port': port.id})
for net_name in nets:
    # check whether network exists
    net = conn.network.find_network(net_name)
    if net:
        log.info('appending network {}:{}'.format(net_name,net.id))
        nics.append({'uuid': net.id})
    else:
        log.warn('net {} not found'.format(net))

conn.compute.create_server(name = name,
                            image_id = im.id,
                            flavor_id = fl.id,
                            security_groups = security_groups,
                            availability_zone = availability_zone,
                            key_name = key,
                            networks = nics,
                            user_data = user_data
                            )

log.info("Sleeping for 5s after create command")
time.sleep(5)

try:
    check_instance = conn.compute.find_server(name)
    while check_instance.status != 'ACTIVE' and check_instance.status != 'ERROR':
        log.info("Building instance... please wait...")
        time.sleep(30)
        check_instance = conn.compute.get_server(check_instance.id)
except:
    raise Exception("Error building instance {}. Probably a nectar outage.".format(check_instance.name))

if check_instance.status == 'ERROR':
    raise Exception("Error building instance {}. Please check OpenStack logs: {}".format(check_instance.name,check_instance.fault))

log.info("VM running... status=" + conn.compute.get_server(check_instance.id).status)

log.info("Instance {}:{} created and running".format(check_instance.name, check_instance.id))
# wait a bit for VM building up
time.sleep(30)
# push instance id and ip address
kwargs["task_instance"].xcom_push(key = "instance-id", value = check_instance.id)
kwargs["task_instance"].xcom_push(
    key="private-ip", value = check_instance.addresses['qld'][0]["addr"]
)

upload taskscripts. skip this step if taskscript is written in user data

grant write permission of the directory for uploading taskscripts.

Code Block

language	py

ssh_hook = SSHHook("ssh_ecoplots")
ssh_hook.username = "core"
grant_write_permission = SSHOperator(
    executor_config=executor_config,
    # The ID specified for the task.
    task_id="grant_write_permission",
    ssh_hook=ssh_hook,
    remote_host="{{ task_instance.xcom_pull(task_ids='create_pdal_instance', key = 'private-ip') }}",
    command=" ".join(["sudo", "chmod", "777", "/var/opt"])
)

upload taskscript

Code Block

language	py

taskscript_upload = SFTPOperator(
    executor_config=executor_config,
    # The ID specified for the task.
    task_id="taskscript_upload",
    ssh_hook=ssh_hook,
    remote_host="{{ task_instance.xcom_pull(task_ids='create_pdal_instance', key = 'private-ip') }}",
    local_filepath = os.path.dirname(os.path.realpath(__file__)) + "/drone_data_update_dag/pdal_task.py",
    remote_filepath="/var/opt/pdal_task.py",
    operation="put",
    create_intermediate_dirs=True
)

run taskscript with VM

Code Block

language	py

pdal_trans = SSHOperator(
    executor_config = executor_config,
    # The ID specified for the task.
    task_id = "pdal_trans",
    ssh_hook = ssh_hook,
    remote_host = "{{ task_instance.xcom_pull(task_ids='create_pdal_instance', key = 'private-ip') }}",
    command = " ".join(["podman", "run", "--rm", "-it",
        "-v", "/var/opt/pdal_task.py:/pdal_task.py",
        "-v", "/var/mnt/data/Q4646:/var/mnt/data/Q4646",
        "--userns=keep-id", "--group-add=keep-groups",
        "--security-opt", "label:disable",
        "docker.io/pdal/pdal:sha-3afba0f9",
        "python3", "/pdal_task.py", *pdal_params])
)

delete vm after use

Code Block

language	py

instance_id = kwargs["task_instance"].xcom_pull(task_ids='create_pdal_instance', key = 'instance-id')
conn = openstack.connect(**auth)
server=conn.compute.get_server(instance_id)
conn.compute.delete_server(server)

...

Versions Compared

Old Version 6

New Version Current

Key