Rework ceph support, new solution use block device mapping; remove privileged celery solution

ceph.py

 import rados
 import rbd
 import os
+import subprocess
 
-from privcelery import celery
+
+DUMP_SIZE_LIMIT = int(os.getenv("DUMP_SIZE_LIMIT", 20 * 1024 ** 3))  # 20GB
 
 
 class CephConnection:
@@ -33,59 +35,50 @@ class CephConnection:
         self.cluster.shutdown()
 
 
-@celery.task
-def write_to_ceph_block_device(poolname, diskname):
+def sudo(*args):
+    try:
+        subprocess.check_output(["/bin/sudo"] + list(args))
+    except subprocess.CalledProcessError as e:
+        raise Exception(e)
+
+
+def map_rbd(ceph_path, local_path):
+    try:
+        sudo("/bin/rbd", "map", ceph_path)
+    except:
+        sudo("/bin/rbd", "unmap", local_path)
+        sudo("/bin/rbd", "map", ceph_path)
+
+
+def save(domain, poolname, diskname):
     diskname = str(diskname)
-    path = "/tmp/" + diskname
-    statinfo = os.stat(path)
-    disk_size = statinfo.st_size
-    with open(path, "rb") as f:
-        with CephConnection(str(poolname)) as conn:
+    poolname = str(poolname)
+    ceph_path = "%s/%s" % (poolname, diskname)
+    local_path = "/dev/rbd/" + ceph_path
+    disk_size = DUMP_SIZE_LIMIT
+
+    with CephConnection(poolname) as conn:
         rbd_inst = rbd.RBD()
         try:
             rbd_inst.create(conn.ioctx, diskname, disk_size)
         except rbd.ImageExists:
             rbd_inst.remove(conn.ioctx, diskname)
             rbd_inst.create(conn.ioctx, diskname, disk_size)
-
         try:
-                with rbd.Image(conn.ioctx, diskname) as image:
-                    offset = 0
-                    data = f.read(4096)
-                    while data:
-                        offset += image.write(data, offset)
-                        data = f.read(4096)
+            map_rbd(ceph_path, local_path)
+            domain.save(local_path)
         except:
             rbd_inst.remove(conn.ioctx, diskname)
             raise
 
 
-@celery.task
-def read_from_ceph_block_device(poolname, diskname):
+def restore(connection, poolname, diskname):
     diskname = str(diskname)
-    path = "/tmp/" + diskname
-    try:
-        with open(path, "wb") as f:
-            with CephConnection(str(poolname)) as conn:
-                with rbd.Image(conn.ioctx, diskname) as image:
-                    offset = 0
-                    size = image.size()
-                    while offset < size - 4096:
-                        data = image.read(offset, 4096)
-                        f.write(data)
-                        offset += 4096
-                    data = image.read(offset, size - offset)
-                    f.write(data)
-                rbd_inst = rbd.RBD()
-                rbd_inst.remove(conn.ioctx, diskname)
-    except:
-        with CephConnection(str(poolname)) as conn:
+    poolname = str(poolname)
+    local_path = "/dev/rbd/%s/%s" % (poolname, diskname)
+
+    connection.restore(local_path)
+    sudo("/bin/rbd", "unmap", local_path)
+    with CephConnection(poolname) as conn:
         rbd_inst = rbd.RBD()
         rbd_inst.remove(conn.ioctx, diskname)
-        remove_temp_file(path)
-        raise
-
-
-@celery.task
-def remove_temp_file(path):
-    os.unlink(path)

miscellaneous/node.conf

@@ -13,11 +13,6 @@ pre-start script
     do
         start netcelery NAME=$hostname.$inst || :
     done
-
-    for inst in vm.priv
-    do
-        start privcelery NAME=$hostname.$inst || :
-    done
 end script
 
 post-stop script
@@ -29,8 +24,4 @@ post-stop script
     do
         stop netcelery NAME=$inst || :
     done
-    for inst in `initctl list|grep "^privcelery "|awk '{print $2}'|tr -d ')'|tr -d '('`
-    do
-        stop privcelery NAME=$inst || :
-    done
 end script

miscellaneous/node.service

@@ -4,7 +4,6 @@ After=network.target
 BindsTo=netcelery@net.fast.service
 BindsTo=vmcelery@vm.fast.service
 BindsTo=vmcelery@vm.slow.service
-BindsTo=privcelery@vm.priv.service
 BindsTo=agentdriver.service
 
 [Service]

miscellaneous/privcelery.conf

-description	"IK Cloud Django Development Server"
-
-respawn
-respawn limit 30 30
-setuid root
-setgid root
-
-instance $NAME
-
-script
-    cd /home/cloud/vmdriver
-    . /home/cloud/.virtualenvs/vmdriver/local/bin/activate
-    . /home/cloud/.virtualenvs/vmdriver/local/bin/postactivate
-    exec celery -A privcelery worker --loglevel=info -n ${NAME}
-end script

miscellaneous/privcelery@.service

-[Unit]
-Description=privcelery %I
-BindsTo=node.service
-
-[Service]
-User=root
-Group=root
-
-KillSignal=SIGTERM
-TimeoutStopSec=600
-Restart=always
-
-WorkingDirectory=/home/cloud/vmdriver
-
-ExecStart=/bin/bash -c "source /etc/profile; source /home/cloud/.virtualenvs/vmdriver/bin/activate; source /home/cloud/.virtualenvs/vmdriver/bin/postactivate; export C_FORCE_ROOT=\"true\"; exec celery -A privcelery worker --loglevel=info -n $(/bin/hostname -s).%I"

privcelery.py

-""" Celery module for privileged RPC calls. """
-from celery import Celery
-from kombu import Queue, Exchange
-from os import getenv
-
-from argparse import ArgumentParser
-
-parser = ArgumentParser()
-parser.add_argument("-n", "--hostname", dest="hostname",
-                    help="Define the full queue name with"
-                    "with priority", metavar="hostname.queue.priority")
-(args, unknwon_args) = parser.parse_known_args()
-HOSTNAME = vars(args).pop("hostname")
-if HOSTNAME is None:
-    raise Exception("You must define hostname as -n <hostname> or "
-                    "--hostname=<hostname>.\n"
-                    "Hostname format must be hostname.module.priority.")
-
-AMQP_URI = getenv('AMQP_URI')
-
-celery = Celery('privcelery',
-                broker=AMQP_URI,
-                include=['ceph'])
-
-celery.conf.update(
-    CELERY_RESULT_BACKEND='amqp',
-    CELERY_TASK_RESULT_EXPIRES=300,
-    CELERY_QUEUES=(
-        Queue(HOSTNAME, Exchange(
-            'ceph', type='direct'), routing_key="ceph"),
-    )
-)

util.py

-import socket
-import logging
-
-
-logger = logging.getLogger(__name__)
-
-
-def get_priviliged_queue_name():
-    hostname = socket.gethostname()
-    logger.debug("Checking for vmdriver priviliged queue %s.vm.priv",
-                 hostname)
-    queue_name = hostname + '.vm.priv'
-    return queue_name

vmdriver.py

@@ -16,10 +16,7 @@ from vm import VMInstance, VMDisk, CephVMDisk, VMNetwork
 
 from vmcelery import celery, lib_connection, to_bool
 
-from ceph import (write_to_ceph_block_device,
-                  read_from_ceph_block_device,
-                  remove_temp_file)
-from util import get_priviliged_queue_name
+import ceph
 
 
 sys.path.append(os.path.dirname(os.path.basename(__file__)))
@@ -334,18 +331,7 @@ def save(name, data_store_type, dir, filename):
     """ Stop virtual machine and save its memory to path. """
     domain = lookupByName(name)
     if data_store_type == "ceph_block":
-        try:
-            path = "/tmp/" + filename
-            domain.save(path)
-            queue_name = get_priviliged_queue_name()
-            write_to_ceph_block_device.apply_async(
-                args=[dir, filename], queue=queue_name).get(timeout=300)
-            remove_temp_file.apply_async(
-                args=[path], queue=queue_name).get(timeout=300)
-        except:
-            remove_temp_file.apply_async(
-                args=[path], queue=queue_name).get(timeout=300)
-            raise
+        ceph.save(domain, dir, filename)
     else:
         path = dir + "/" + filename
         domain.save(path)
@@ -363,13 +349,7 @@ def restore(name, data_store_type, dir, filename):
 
     """
     if data_store_type == "ceph_block":
-        path = "/tmp/" + filename
-        queue_name = get_priviliged_queue_name()
-        read_from_ceph_block_device.apply_async(
-            args=[dir, filename], queue=queue_name).get(timeout=300)
-        Connection.get().restore(path)
-        remove_temp_file.apply_async(
-            args=[path], queue=queue_name).get(timeout=300)
+        ceph.restore(Connection.get(), dir, filename)
     else:
         path = dir + "/" + filename
         Connection.get().restore(path)