rpm/libvirt
SHA256
1
0
Fork 0
forked from pool/libvirt
Code Pull Requests Activity

Accepting request 563669 from home:jfehlig:branches:Virtualization

Browse Source 
- Add a qemu hook script providing functionality similar to Xen's
  block-dmmd script
  suse-qemu-domain-hook.py
  FATE#324177

OBS-URL: https://build.opensuse.org/request/show/563669
OBS-URL: https://build.opensuse.org/package/show/Virtualization/libvirt?expand=0&rev=659
This commit is contained in:
James Fehlig 2018-01-11 18:22:07 +00:00 committed by Git OBS Bridge
parent 18873bd642
commit 6588d246d6
3 changed files with 355 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
libvirt.changes
View File

@ -1,3 +1,11 @@
-------------------------------------------------------------------
Thu Jan 11 16:52:43 UTC 2018 - jfehlig@suse.com
- Add a qemu hook script providing functionality similar to Xen's
block-dmmd script
suse-qemu-domain-hook.py
FATE#324177
-------------------------------------------------------------------
Wed Jan 3 10:46:26 UTC 2018 - cbosdonnat@suse.com

16
libvirt.spec
View File

@ -305,6 +305,7 @@ Source1: %{name}-%{version}.tar.xz.asc
Source2: %{name}.keyring
Source3: libvirtd-relocation-server.fw
Source4: libvirt-supportconfig
Source5: suse-qemu-domain-hook.py
Source99: baselibs.conf
Source100: %{name}-rpmlintrc
# Upstream patches
@ -407,6 +408,15 @@ Server side daemon required to manage the virtualization capabilities
of recent versions of Linux. Requires a hypervisor specific sub-RPM
for specific drivers.
%package daemon-hooks
Summary: Hook scripts for the libvirtd daemon
Group: Development/Libraries/C and C++
Requires: %{name}-daemon = %{version}-%{release}
%description daemon-hooks
Default hook scripts for the libvirt daemon
%package daemon-config-network
Summary: Default configuration files for the libvirtd daemon
Group: Development/Libraries/C and C++
@ -1233,6 +1243,9 @@ install -m 644 %{S:3} %{buildroot}/%{_fwdefdir}/libvirtd-relocation-server
mkdir -p %{buildroot}/usr/lib/supportconfig/plugins
install -m 755 %{S:4} %{buildroot}/usr/lib/supportconfig/plugins/libvirt
# install qemu hook script
install -m 755 %{S:5} %{buildroot}/%{_sysconfdir}/libvirt/hooks/qemu
%ifarch %{power64} s390x x86_64
mv %{buildroot}/%{_datadir}/systemtap/tapset/libvirt_probes.stp \
%{buildroot}/%{_datadir}/systemtap/tapset/libvirt_probes-64.stp
@ -1414,6 +1427,9 @@ fi
%dir /usr/lib/supportconfig/plugins
/usr/lib/supportconfig/plugins/libvirt
%files daemon-hooks
%{_sysconfdir}/libvirt/hooks/qemu
%files daemon-config-network
%dir %{_datadir}/libvirt/networks/
%{_datadir}/libvirt/networks/default.xml

331
suse-qemu-domain-hook.py Normal file
View File

@ -0,0 +1,331 @@
#!/usr/bin/env python
# libvirt hook script for QEMU/KVM domains. See the libvirt hooks
# documenation for more details
#
# https://www.libvirt.org/hooks.html
#
# Currently this hook looks for domains with <metadata> containing
# configuration for dmmd <disk> devices. All <metadata> sub-elements
# must have a matching <disk> under <devices>. Those without a matching
# <disk> will be ignored.
#
# The dmmd device syntax is similar to Xen's block-dmmd. E.g.
# md;/dev/md0(/etc/mdadm/mdadm.conf);lvm;/dev/vg/lv
#
# Device pairs (type;dev) are processed in order. The last device
# should match a <source dev=> attribute of a <disk>. The following
# configuration illustrates a domain with two dmmd devices
#
# <domain>
# ...
# <metadata>
# <hook:dmmd xmlns:hook='https://libvirt.org/schemas/domain/hooks/1.0'>
# <disk>md;/dev/md0(/etc/mdadm.conf);lvm;/dev/vg1/lv1</disk>
# <disk>md;/dev/md1(/etc/mdadm.conf);lvm;/dev/vg1/lv2</disk>
# </hook:dmmd>
# </metadata>
# <devices>
# ...
# <disk type='block' device='disk'>
# <driver name='qemu' type='raw'/>
# <source dev='/dev/vg1/lv1'/>
# <target dev='vdb' bus='virtio'/>
# </disk>
# <disk type='block' device='disk'>
# <driver name='qemu' type='raw'/>
# <source dev='/dev/vg1/lv2'/>
# <target dev='vdc' bus='virtio'/>
# </disk>
# </devices>
# </domain>
#
#
# md devices can optionally:
# specify a config file through:
# md;/dev/md100(/var/opt/config/mdadm.conf)
# use an array name (mdadm -N option):
# md;My-MD-name;lvm;/dev/vg1/lv1
import os
import sys
import time
import subprocess
from lxml import etree
from subprocess import check_output
from subprocess import CalledProcessError
COMMAND_TIMEOUT = 60
MDADM_BIN = "/sbin/mdadm"
PVSCAN_BIN = "/sbin/pvscan"
LVCHANGE_BIN = "/sbin/lvchange"
HOOK_NAMESPACE = "https://libvirt.org/schemas/domain/hooks/1.0"
HOOK_NS_TAG = "{%s}" % HOOK_NAMESPACE
DEBUG = False
def write_debug(msg):
if DEBUG:
with open("/var/log/libvirt/qemu/suse-qemu-hook-output.log", "a") as f:
f.write(msg + "\n")
def run_cmd(cmd):
cmd_output = ""
rc = 0
msg = ""
for m in cmd:
msg += m + " "
write_debug("run_cmd executing: " + msg)
try:
cmd_output = check_output(cmd, stderr=subprocess.STDOUT)
except CalledProcessError as err:
write_debug("run_cmd: caught CalledProcessError with output: " + err.output)
rc = err.returncode
if rc != 0:
write_debug("run_cmd failed: " + msg)
return [rc, cmd_output]
def prepare_md(dev):
conf = []
mdadmopts = []
devpath = ""
startcfg = dev.find("(")
# check if MD config specifies a conf file for mdadm
if startcfg != -1:
endcfg = dev.find(")")
conf = ["-c"]
conf.append(dev[startcfg + 1:endcfg])
dev = dev[:startcfg]
# check if MD config contains a device or array name
if not dev.startswith("/"):
mdadmopts = ["-s"]
mdadmopts.append("-N")
devpath = "/dev/md/" + dev
else:
devpath = dev
# check if MD device is already active
cmd = [MDADM_BIN, "-Q"]
cmd.append(devpath)
write_debug("prepare_md: calling mdadm -Q for device " + devpath)
ret, cmd_output = run_cmd(cmd)
if ret == 0:
write_debug("prepare_md: mdadm -Q succeeded for device " + devpath + ". Already activated")
return 0
cmd = [MDADM_BIN, "-A"]
cmd.extend(mdadmopts)
cmd.extend(conf)
cmd.append(devpath)
write_debug("prepare_md: calling mdadm -A for device " + devpath)
ret, cmd_output = run_cmd(cmd)
if ret != 0:
write_debug("prepare_md: mdadm -A failed for device " + devpath)
else:
write_debug("prepare_md: mdadm -A succeeded for device " + devpath)
return ret
def release_md(dev):
conf = []
devpath = ""
startcfg = dev.find("(")
if startcfg != -1:
endcfg = dev.find(")")
conf = ["-c"]
conf.append(dev[startcfg + 1:endcfg])
dev = dev[:startcfg]
# check if MD config contains a device or array name. For
# querying and deactivating a device name is required
if not dev.startswith("/"):
devpath = "/dev/md/" + dev
else:
devpath = dev
# check if device exists
cmd = [MDADM_BIN, "-Q"]
cmd.extend(conf)
cmd.append(devpath)
write_debug("release_md: calling mdadm -Q for device " + devpath)
ret, cmd_output = run_cmd(cmd)
if ret != 0:
write_debug("release_md: mdadm -Q failed for device " + devpath + ". Already deactivated")
return 0
cmd = [MDADM_BIN, "-S"]
cmd.extend(conf)
cmd.append(devpath)
write_debug("release_md: calling mdadm -S for device " + devpath)
ret, cmd_output = run_cmd(cmd)
if ret == 0:
write_debug("release_md: mdadm -S succeeded for device " + devpath)
else:
write_debug("release_md: mdadm -S failed for device " + devpath)
return ret
def prepare_lvm(dev):
cmd = [LVCHANGE_BIN]
cmd.append("-aey")
cmd.append(dev)
endtime = time.time() + COMMAND_TIMEOUT;
while time.time() < endtime:
# When using MD devices for LVM PV, it is best to rescan for PV and VG
run_cmd([PVSCAN_BIN])
ret, cmd_output = run_cmd(cmd)
if ret == 0 and os.path.exists(dev):
write_debug("prepare_lvm: lvchange -aey succeeded on device " + dev)
return 0
else:
write_debug("prepare_lvm: lvchange -aey failed on device " + dev)
time.sleep(0.1)
write_debug("prepare_lvm: lvchange -aey never succeeded for device " + dev)
return 1
def release_lvm(dev):
# Nothing to do if the device doesn't exist or is already deactivated
if not os.path.exists(dev):
write_debug("release_lvm: dev " + dev + " does not exist. Nothing to do!")
return 0
cmd = [LVCHANGE_BIN]
cmd.append("-aen")
cmd.append(dev)
endtime = time.time() + COMMAND_TIMEOUT;
while time.time() < endtime:
ret, cmd_output = run_cmd(cmd)
if ret == 0:
write_debug("release_lvm: lvchange -aen succeeded for device " + dev)
return 0
else:
write_debug("release_lvm: lvchange -aen failed for device " + dev + ". Trying again...")
time.sleep(0.1)
write_debug("release_lvm: lvchange -aen never succeeded for device " + dev)
return 1
def prepare_config(params):
write_debug("prepare_config: called with params " + params)
conf = params.split(";")
i = 0
while i < len(conf):
t = conf[i]
d = conf[i+1]
write_debug("prepare_config: got t = " + t + " and d = " + d)
if t == "md":
if prepare_md(d):
write_debug("prepare_config: failed to prepare MD device " + d)
return 1
if t == "lvm":
if prepare_lvm(d):
write_debug("prepare_config: failed to prepare LVM device " + d)
return 1
i += 2
return 0
def release_config(params):
write_debug("release_config: called with params " + params)
conf = params.split(";")
i = len(conf) - 1
ret = 0
# work backwards through the list when releasing, cleaning
# up LVM first, then MD
while i >= 0:
t = conf[i-1]
d = conf[i]
write_debug("release_config: got t = " + t + " and d = " + d)
if t == "md":
if release_md(d):
write_debug("release_config: failed to release MD device " + d)
ret = 1
if t == "lvm":
if release_lvm(d):
write_debug("release_config: failed to release LVM device " + d)
ret = 1
i -= 2
return ret
if len(sys.argv) < 3:
sys.exit(1)
exit_code = 0
disk_devs = []
phase = sys.argv[2]
vmxml = sys.stdin.read()
tree = etree.fromstring(vmxml)
devs = tree.xpath("/domain/devices/disk")
dmmd_configs = tree.xpath("/domain/metadata/hook:dmmd/disk", namespaces={'hook': HOOK_NAMESPACE})
if len(dmmd_configs) == 0:
write_debug("No dmmd configurations found in <metadata>")
sys.exit(0)
write_debug("got phase: " + phase)
# build a list of <disk type='block'> source device names to check against
# dmmd configurations
for d in devs:
val = d.get("type")
if val is None or val != "block":
continue
for child in d:
if child.tag == "source":
disk_devs.append(child.get("dev"))
# For each dmmd configuration in <metadata>, check there is a corresponding
# disk
for config in dmmd_configs:
# check that a disk exists for this config. <disk> devices may have
# been added or removed without a corresponding update to <metadata>
index = config.text.rfind(";")
if index == -1:
continue
disk = config.text[index + 1:]
# remove config file specfied with '(/path/to/conf)'
if disk.endswith(")"):
index = disk.rfind("(")
if index == -1:
continue
disk = disk[:index]
if disk not in disk_devs:
write_debug("Ignoring config '" + config.text + "' with no matching <disk> device")
continue
# TODO: check that migration can be handled by the 'prepare' phase on
# destination and 'release' phase on source
if phase == "prepare":
exit_code = prepare_config(config.text)
if phase == "release":
exit_code = release_config(config.text)
sys.exit(exit_code)