blob: 12d8eefd2fd9c6536b816d2f6e95d60456c9b252 [file] [log] [blame]
# Copyright 2009 Google Inc. Released under the GPL v2
This module defines the base classes for the Host hierarchy.
Implementation details:
You should import the "hosts" package instead of importing each type of host.
Host: a machine on which you can run programs
__author__ = """ (Martin J. Bligh), (Benjamin Poirier), (Ryan Stutsman)
import cPickle, cStringIO, logging, os, re, time
from autotest_lib.client.common_lib import global_config, error, utils
from autotest_lib.client.common_lib import host_protections
from autotest_lib.client.common_lib.cros.graphite import autotest_stats
from autotest_lib.client.bin import partition
class Host(object):
This class represents a machine on which you can run programs.
It may be a local machine, the one autoserv is running on, a remote
machine or a virtual machine.
Implementation details:
This is an abstract class, leaf subclasses must implement the methods
listed here. You must not instantiate this class but should
instantiate one of those leaf subclasses.
When overriding methods that raise NotImplementedError, the leaf class
is fully responsible for the implementation and should not chain calls
to super. When overriding methods that are a NOP in Host, the subclass
should chain calls to super(). The criteria for fitting a new method into
one category or the other should be:
1. If two separate generic implementations could reasonably be
concatenated, then the abstract implementation should pass and
subclasses should chain calls to super.
2. If only one class could reasonably perform the stated function
(e.g. two separate run() implementations cannot both be executed)
then the method should raise NotImplementedError in Host, and
the implementor should NOT chain calls to super, to ensure that
only one implementation ever gets executed.
job = None
DEFAULT_REBOOT_TIMEOUT = global_config.global_config.get_config_value(
"HOSTS", "default_reboot_timeout", type=int, default=1800)
WAIT_DOWN_REBOOT_TIMEOUT = global_config.global_config.get_config_value(
"HOSTS", "wait_down_reboot_timeout", type=int, default=840)
WAIT_DOWN_REBOOT_WARNING = global_config.global_config.get_config_value(
"HOSTS", "wait_down_reboot_warning", type=int, default=540)
HOURS_TO_WAIT_FOR_RECOVERY = global_config.global_config.get_config_value(
"HOSTS", "hours_to_wait_for_recovery", type=float, default=2.5)
# the number of hardware repair requests that need to happen before we
# actually send machines to hardware repair
OP_REBOOT = 'reboot'
OP_SUSPEND = 'suspend'
def __init__(self, *args, **dargs):
self._initialize(*args, **dargs)
def _initialize(self, *args, **dargs):
self._already_repaired = []
self._removed_files = False
def close(self):
def setup(self):
def run(self, command, timeout=3600, ignore_status=False,
stdout_tee=utils.TEE_TO_LOGS, stderr_tee=utils.TEE_TO_LOGS,
stdin=None, args=()):
Run a command on this host.
@param command: the command line string
@param timeout: time limit in seconds before attempting to
kill the running process. The run() function
will take a few seconds longer than 'timeout'
to complete if it has to kill the process.
@param ignore_status: do not raise an exception, no matter
what the exit code of the command is.
@param stdout_tee/stderr_tee: where to tee the stdout/stderr
@param stdin: stdin to pass (a string) to the executed command
@param args: sequence of strings to pass as arguments to command by
quoting them in " and escaping their contents if necessary
@return a utils.CmdResult object
@raises AutotestHostRunError: the exit code of the command execution
was not 0 and ignore_status was not enabled
raise NotImplementedError('Run not implemented!')
def run_output(self, command, *args, **dargs):
return, *args, **dargs).stdout.rstrip()
def reboot(self):
raise NotImplementedError('Reboot not implemented!')
def suspend(self):
raise NotImplementedError('Suspend not implemented!')
def sysrq_reboot(self):
raise NotImplementedError('Sysrq reboot not implemented!')
def reboot_setup(self, *args, **dargs):
def reboot_followup(self, *args, **dargs):
def get_file(self, source, dest, delete_dest=False):
raise NotImplementedError('Get file not implemented!')
def send_file(self, source, dest, delete_dest=False):
raise NotImplementedError('Send file not implemented!')
def get_tmp_dir(self):
raise NotImplementedError('Get temp dir not implemented!')
def is_up(self):
raise NotImplementedError('Is up not implemented!')
def is_shutting_down(self):
""" Indicates is a machine is currently shutting down. """
# runlevel() may not be available, so wrap it in try block.
runlevel = int("runlevel").stdout.strip().split()[1])
return runlevel in (0, 6)
return False
def get_wait_up_processes(self):
""" Gets the list of local processes to wait for in wait_up. """
get_config = global_config.global_config.get_config_value
proc_list = get_config("HOSTS", "wait_up_processes",
processes = set(p.strip() for p in proc_list.split(","))
return processes
def get_boot_id(self, timeout=60):
""" Get a unique ID associated with the current boot.
Should return a string with the semantics such that two separate
calls to Host.get_boot_id() return the same string if the host did
not reboot between the two calls, and two different strings if it
has rebooted at least once between the two calls.
@param timeout The number of seconds to wait before timing out.
@return A string unique to this boot or None if not available."""
BOOT_ID_FILE = '/proc/sys/kernel/random/boot_id'
NO_ID_MSG = 'no boot_id available'
cmd = 'if [ -f %r ]; then cat %r; else echo %r; fi' % (
boot_id =, timeout=timeout).stdout.strip()
if boot_id == NO_ID_MSG:
return None
return boot_id
def wait_up(self, timeout=None):
raise NotImplementedError('Wait up not implemented!')
def wait_down(self, timeout=None, warning_timer=None, old_boot_id=None):
raise NotImplementedError('Wait down not implemented!')
def construct_host_metadata(self, type_str):
"""Returns dict of metadata with type_str, hostname, time_recorded.
@param type_str: String representing _type field in es db.
For example: type_str='reboot_total'.
metadata = {
'hostname': self.hostname,
'time_recorded': time.time(),
'_type': type_str,
return metadata
def wait_for_restart(self, timeout=DEFAULT_REBOOT_TIMEOUT,
log_failure=True, old_boot_id=None, **dargs):
""" Wait for the host to come back from a reboot. This is a generic
implementation based entirely on wait_up and wait_down. """
key_string = 'Reboot.%s' % dargs.get('board')
total_reboot_timer = autotest_stats.Timer('' % key_string,
wait_down_timer = autotest_stats.Timer('%s.wait_down' % key_string,
if not self.wait_down(timeout=down_timeout,
if log_failure:
self.record("ABORT", None, "reboot.verify", "shut down failed")
raise error.AutoservShutdownError("Host did not shut down")
wait_up_timer = autotest_stats.Timer('%s.wait_up' % key_string,
if self.wait_up(timeout):
self.record("GOOD", None, "reboot.verify")
self.record("ABORT", None, "reboot.verify",
"Host did not return from reboot")
raise error.AutoservRebootError("Host did not return from reboot")
def verify(self):
def verify_hardware(self):
def verify_connectivity(self):
def verify_software(self):
def check_diskspace(self, path, gb):
"""Raises an error if path does not have at least gb GB free.
@param path The path to check for free disk space.
@param gb A floating point number to compare with a granularity
of 1 MB.
1000 based SI units are used.
@raises AutoservDiskFullHostError if path has less than gb GB free.
one_mb = 10 ** 6 # Bytes (SI unit).
mb_per_gb = 1000.0'Checking for >= %s GB of space under %s on machine %s',
gb, path, self.hostname)
df ='df -PB %d %s | tail -1' % (one_mb, path)).stdout.split()
free_space_gb = int(df[3]) / mb_per_gb
if free_space_gb < gb:
raise error.AutoservDiskFullHostError(path, gb, free_space_gb)
else:'Found %s GB >= %s GB of space under %s on machine %s',
free_space_gb, gb, path, self.hostname)
def check_inodes(self, path, min_kilo_inodes):
"""Raises an error if a file system is short on i-nodes.
@param path The path to check for free i-nodes.
@param min_kilo_inodes Minimum number of i-nodes required,
in units of 1000 i-nodes.
@raises AutoservNoFreeInodesError If the minimum required
i-node count isn't available.
min_inodes = 1000 * min_kilo_inodes'Checking for >= %d i-nodes under %s '
'on machine %s', min_inodes, path, self.hostname)
df ='df -Pi %s | tail -1' % path).stdout.split()
free_inodes = int(df[3])
if free_inodes < min_inodes:
raise error.AutoservNoFreeInodesError(path, min_inodes,
else:'Found %d >= %d i-nodes under %s on '
'machine %s', free_inodes, min_inodes,
path, self.hostname)
def get_open_func(self, use_cache=True):
Defines and returns a function that may be used instead of built-in
open() to open and read files. The returned function is implemented
by using'cat <file>') and may cache the results for the same
@param use_cache Cache results of'cat <filename>') for the
same filename
@return a function that can be used instead of built-in open()
cached_files = {}
def open_func(filename):
if not use_cache or filename not in cached_files:
output ='cat \'%s\'' % filename,
stdout_tee=open('/dev/null', 'w')).stdout
fd = cStringIO.StringIO(output)
if not use_cache:
return fd
cached_files[filename] = fd
return cached_files[filename]
return open_func
def check_partitions(self, root_part, filter_func=None):
""" Compare the contents of /proc/partitions with those of
/proc/mounts and raise exception in case unmounted partitions are found
root_part: in Linux /proc/mounts will never directly mention the root
partition as being mounted on / instead it will say that /dev/root is
mounted on /. Thus require this argument to filter out the root_part
from the ones checked to be mounted
filter_func: unnary predicate for additional filtering out of
partitions required to be mounted
Raise: error.AutoservHostError if unfiltered unmounted partition found
print 'Checking if non-swap partitions are mounted...'
unmounted = partition.get_unmounted_partition_list(root_part,
filter_func=filter_func, open_func=self.get_open_func())
if unmounted:
raise error.AutoservNotMountedHostError(
'Found unmounted partitions: %s' %
[part.device for part in unmounted])
def _repair_wait_for_reboot(self):
if self.is_shutting_down():'Host is shutting down, waiting for a restart')
def _get_mountpoint(self, path):
"""Given a "path" get the mount point of the filesystem containing
that path."""
code = ('import os\n'
# sanitize the path and resolve symlinks
'path = os.path.realpath(%r)\n'
"while path != '/' and not os.path.ismount(path):\n"
' path, _ = os.path.split(path)\n'
'print path\n') % path
return'python -c "%s"' % code,
stdout_tee=open(os.devnull, 'w')).stdout.rstrip()
def erase_dir_contents(self, path, ignore_status=True, timeout=3600):
"""Empty a given directory path contents."""
rm_cmd = 'find "%s" -mindepth 1 -maxdepth 1 -print0 | xargs -0 rm -rf' % path, ignore_status=ignore_status, timeout=timeout)
self._removed_files = True
def repair_full_disk(self, mountpoint):
# it's safe to remove /tmp and /var/tmp, site specific overrides may
# want to remove some other places too
if mountpoint == self._get_mountpoint('/tmp'):
if mountpoint == self._get_mountpoint('/var/tmp'):
def _call_repair_func(self, err, func, *args, **dargs):
for old_call in self._already_repaired:
if old_call == (func, args, dargs):
# re-raising the original exception because surrounding
# error handling may want to try other ways to fix it
logging.warning('Already done this (%s) repair procedure, '
're-raising the original exception.', func)
raise err
func(*args, **dargs)
except (error.AutoservHardwareRepairRequestedError,
# let these special exceptions propagate
except error.AutoservError:
logging.exception('Repair failed but continuing in case it managed'
' to repair enough')
self._already_repaired.append((func, args, dargs))
def repair_filesystem_only(self):
"""perform file system repairs only"""
while True:
# try to repair specific problems
try:'Running verify to find failures to repair...')
if self._removed_files:'Removed files, rebooting to release the'
' inodes')
return # verify succeeded, then repair succeeded
except error.AutoservHostIsShuttingDownError, err:
logging.exception('verify failed')
self._call_repair_func(err, self._repair_wait_for_reboot)
except error.AutoservDiskFullHostError, err:
logging.exception('verify failed')
self._call_repair_func(err, self.repair_full_disk,
def repair_software_only(self):
"""perform software repairs only"""
while True:
except (error.AutoservSshPingHostError, error.AutoservSSHTimeout,
error.AutoservDiskFullHostError), err:
logging.exception('verify failed')'Trying to reinstall the machine')
self._call_repair_func(err, self.machine_install)
def repair_full(self):
hardware_repair_requests = 0
while True:
except error.AutoservHardwareRepairRequiredError, err:
logging.exception('software repair failed, '
'hardware repair requested')
hardware_repair_requests += 1
try_hardware_repair = (hardware_repair_requests >=
if try_hardware_repair:'hardware repair requested %d times, '
'trying hardware repair',
self._call_repair_func(err, self.request_hardware_repair)
else:'hardware repair requested %d times, '
'trying software repair again',
except error.AutoservHardwareHostError, err:
logging.exception('verify failed')
# software repair failed, try hardware repair'Hardware problem found, '
'requesting hardware repairs')
self._call_repair_func(err, self.request_hardware_repair)
def repair_with_protection(self, protection_level):
"""Perform the maximal amount of repair within the specified
protection level.
@param protection_level: the protection level to use for limiting
repairs, a host_protections.Protection
protection = host_protections.Protection
if protection_level == protection.DO_NOT_REPAIR:'Protection is "Do not repair" so just verifying')
elif protection_level == protection.REPAIR_FILESYSTEM_ONLY:'Attempting filesystem-only repair')
elif protection_level == protection.REPAIR_SOFTWARE_ONLY:'Attempting software repair only')
elif protection_level == protection.NO_PROTECTION:'Attempting full repair')
raise NotImplementedError('Unknown host protection level %s'
% protection_level)
def disable_ipfilters(self):
"""Allow all network packets in and out of the host."""'iptables-save > /tmp/iptable-rules')'iptables -P INPUT ACCEPT')'iptables -P FORWARD ACCEPT')'iptables -P OUTPUT ACCEPT')
def enable_ipfilters(self):
"""Re-enable the IP filters disabled from disable_ipfilters()"""
if self.path_exists('/tmp/iptable-rules'):'iptables-restore < /tmp/iptable-rules')
def cleanup(self):
def machine_install(self):
raise NotImplementedError('Machine install not implemented!')
def install(self, installableObject):
def get_autodir(self):
raise NotImplementedError('Get autodir not implemented!')
def set_autodir(self):
raise NotImplementedError('Set autodir not implemented!')
def start_loggers(self):
""" Called to start continuous host logging. """
def stop_loggers(self):
""" Called to stop continuous host logging. """
# some extra methods simplify the retrieval of information about the
# Host machine, with generic implementations based on run(). subclasses
# should feel free to override these if they can provide better
# implementations for their specific Host types
def get_num_cpu(self):
""" Get the number of CPUs in the host according to /proc/cpuinfo. """
proc_cpuinfo ='cat /proc/cpuinfo',
stdout_tee=open(os.devnull, 'w')).stdout
cpus = 0
for line in proc_cpuinfo.splitlines():
if line.startswith('processor'):
cpus += 1
return cpus
def get_arch(self):
""" Get the hardware architecture of the remote machine. """
arch ='/bin/uname -m').stdout.rstrip()
if re.match(r'i\d86$', arch):
arch = 'i386'
return arch
def get_kernel_ver(self):
""" Get the kernel version of the remote machine. """
return'/bin/uname -r').stdout.rstrip()
def get_cmdline(self):
""" Get the kernel command line of the remote machine. """
return'cat /proc/cmdline').stdout.rstrip()
def get_meminfo(self):
""" Get the kernel memory info (/proc/meminfo) of the remote machine
and return a dictionary mapping the various statistics. """
meminfo_dict = {}
meminfo ='cat /proc/meminfo').stdout.splitlines()
for key, val in (line.split(':', 1) for line in meminfo):
meminfo_dict[key.strip()] = val.strip()
return meminfo_dict
def path_exists(self, path):
""" Determine if path exists on the remote machine. """
result ='ls "%s" > /dev/null' % utils.sh_escape(path),
return result.exit_status == 0
# some extra helpers for doing job-related operations
def record(self, *args, **dargs):
""" Helper method for recording status logs against Host.job that
silently becomes a NOP if Host.job is not available. The args and
dargs are passed on to Host.job.record unchanged. """
if self.job:
self.job.record(*args, **dargs)
def log_kernel(self):
""" Helper method for logging kernel information into the status logs.
Intended for cases where the "current" kernel is not really defined
and we want to explicitly log it. Does nothing if this host isn't
actually associated with a job. """
if self.job:
kernel = self.get_kernel_ver()
self.job.record("INFO", None, None,
optional_fields={"kernel": kernel})
def log_op(self, op, op_func):
""" Decorator for wrapping a management operaiton in a group for status
logging purposes.
@param op: name of the operation.
@param op_func: a function that carries out the operation
(reboot, suspend)
if self.job and not hasattr(self, "RUNNING_LOG_OP"):
self.RUNNING_LOG_OP = True
self.job.run_op(op, op_func, self.get_kernel_ver)
def request_hardware_repair(self):
""" Should somehow request (send a mail?) for hardware repairs on
this machine. The implementation can either return by raising the
special error.AutoservHardwareRepairRequestedError exception or can
try to wait until the machine is repaired and then return normally.
raise NotImplementedError("request_hardware_repair not implemented")
def list_files_glob(self, glob):
Get a list of files on a remote host given a glob pattern path.
SCRIPT = ("python -c 'import cPickle, glob, sys;"
"cPickle.dump(glob.glob(sys.argv[1]), sys.stdout, 0)'")
output =, args=(glob,), stdout_tee=None,
return cPickle.loads(output)
def symlink_closure(self, paths):
Given a sequence of path strings, return the set of all paths that
can be reached from the initial set by following symlinks.
@param paths: sequence of path strings.
@return: a sequence of path strings that are all the unique paths that
can be reached from the given ones after following symlinks.
SCRIPT = ("python -c 'import cPickle, os, sys\n"
"paths = cPickle.load(sys.stdin)\n"
"closure = {}\n"
"while paths:\n"
" path = paths.keys()[0]\n"
" del paths[path]\n"
" if not os.path.exists(path):\n"
" continue\n"
" closure[path] = None\n"
" if os.path.islink(path):\n"
" link_to = os.path.join(os.path.dirname(path),\n"
" os.readlink(path))\n"
" if link_to not in closure.keys():\n"
" paths[link_to] = None\n"
"cPickle.dump(closure.keys(), sys.stdout, 0)'")
input_data = cPickle.dumps(dict((path, None) for path in paths), 0)
output =, stdout_tee=None, stdin=input_data,
return cPickle.loads(output)
def cleanup_kernels(self, boot_dir='/boot'):
Remove any kernel image and associated files (vmlinux,,
modules) for any image found in the boot directory that is not
referenced by entries in the bootloader configuration.
@param boot_dir: boot directory path string, default '/boot'
# find all the vmlinuz images referenced by the bootloader
vmlinuz_prefix = os.path.join(boot_dir, 'vmlinuz-')
boot_info = self.bootloader.get_entries()
used_kernver = [boot['kernel'][len(vmlinuz_prefix):]
for boot in boot_info.itervalues()]
# find all the unused vmlinuz images in /boot
all_vmlinuz = self.list_files_glob(vmlinuz_prefix + '*')
used_vmlinuz = self.symlink_closure(vmlinuz_prefix + kernver
for kernver in used_kernver)
unused_vmlinuz = set(all_vmlinuz) - set(used_vmlinuz)
# find all the unused vmlinux images in /boot
vmlinux_prefix = os.path.join(boot_dir, 'vmlinux-')
all_vmlinux = self.list_files_glob(vmlinux_prefix + '*')
used_vmlinux = self.symlink_closure(vmlinux_prefix + kernver
for kernver in used_kernver)
unused_vmlinux = set(all_vmlinux) - set(used_vmlinux)
# find all the unused files in /boot
systemmap_prefix = os.path.join(boot_dir, '')
all_system_map = self.list_files_glob(systemmap_prefix + '*')
used_system_map = self.symlink_closure(
systemmap_prefix + kernver for kernver in used_kernver)
unused_system_map = set(all_system_map) - set(used_system_map)
# find all the module directories associated with unused kernels
modules_prefix = '/lib/modules/'
all_moddirs = [dir for dir in self.list_files_glob(modules_prefix + '*')
if re.match(modules_prefix + r'\d+\.\d+\.\d+.*', dir)]
used_moddirs = self.symlink_closure(modules_prefix + kernver
for kernver in used_kernver)
unused_moddirs = set(all_moddirs) - set(used_moddirs)
# remove all the vmlinuz files we don't use
# TODO: if needed this should become package manager agnostic
for vmlinuz in unused_vmlinuz:
# try and get an rpm package name
rpm ='rpm -qf', args=(vmlinuz,),
ignore_status=True, timeout=120)
if rpm.exit_status == 0:
packages = set(line.strip() for line in
# if we found some package names, try to remove them
for package in packages:'rpm -e', args=(package,),
ignore_status=True, timeout=120)
# remove the image files anyway, even if rpm didn't'rm -f', args=(vmlinuz,),
ignore_status=True, timeout=120)
# remove all the vmlinux and files left over
for f in (unused_vmlinux | unused_system_map):'rm -f', args=(f,),
ignore_status=True, timeout=120)
# remove all unused module directories
# the regex match should keep us safe from removing the wrong files
for moddir in unused_moddirs:'rm -fr', args=(moddir,), ignore_status=True)