abstract_ssh.py revision 04be2bd5e4666a5c253e9c30ab20555e04286032
1import os, time, socket, shutil, glob, logging, traceback, tempfile
2from autotest_lib.client.common_lib import autotemp, error
3from autotest_lib.server import utils, autotest
4from autotest_lib.server.hosts import remote
5from autotest_lib.client.common_lib.global_config import global_config
6
7# pylint: disable-msg=C0111
8
9get_value = global_config.get_config_value
10enable_master_ssh = get_value('AUTOSERV', 'enable_master_ssh', type=bool,
11                              default=False)
12
13
14class AbstractSSHHost(remote.RemoteHost):
15    """
16    This class represents a generic implementation of most of the
17    framework necessary for controlling a host via ssh. It implements
18    almost all of the abstract Host methods, except for the core
19    Host.run method.
20    """
21
22    def _initialize(self, hostname, user="root", port=22, password="",
23                    *args, **dargs):
24        super(AbstractSSHHost, self)._initialize(hostname=hostname,
25                                                 *args, **dargs)
26        self.ip = socket.getaddrinfo(self.hostname, None)[0][4][0]
27        self.user = user
28        self.port = port
29        self.password = password
30        self._use_rsync = None
31        self.known_hosts_file = tempfile.mkstemp()[1]
32
33        """
34        Master SSH connection background job, socket temp directory and socket
35        control path option. If master-SSH is enabled, these fields will be
36        initialized by start_master_ssh when a new SSH connection is initiated.
37        """
38        self.master_ssh_job = None
39        self.master_ssh_tempdir = None
40        self.master_ssh_option = ''
41
42
43    def make_ssh_command(self, user="root", port=22, opts='',
44                         hosts_file='/dev/null',
45                         connect_timeout=30, alive_interval=300):
46        base_command = ("/usr/bin/ssh -a -x %s -o StrictHostKeyChecking=no "
47                        "-o UserKnownHostsFile=%s -o BatchMode=yes "
48                        "-o ConnectTimeout=%d -o ServerAliveInterval=%d "
49                        "-l %s -p %d")
50        assert isinstance(connect_timeout, (int, long))
51        assert connect_timeout > 0 # can't disable the timeout
52        return base_command % (opts, hosts_file, connect_timeout,
53                               alive_interval, user, port)
54
55
56    def use_rsync(self):
57        if self._use_rsync is not None:
58            return self._use_rsync
59
60        # Check if rsync is available on the remote host. If it's not,
61        # don't try to use it for any future file transfers.
62        self._use_rsync = self._check_rsync()
63        if not self._use_rsync:
64            logging.warning("rsync not available on remote host %s -- disabled",
65                         self.hostname)
66        return self._use_rsync
67
68
69    def _check_rsync(self):
70        """
71        Check if rsync is available on the remote host.
72        """
73        try:
74            self.run("rsync --version", stdout_tee=None, stderr_tee=None)
75        except error.AutoservRunError:
76            return False
77        return True
78
79
80    def _encode_remote_paths(self, paths, escape=True):
81        """
82        Given a list of file paths, encodes it as a single remote path, in
83        the style used by rsync and scp.
84        """
85        if escape:
86            paths = [utils.scp_remote_escape(path) for path in paths]
87        return '%s@%s:"%s"' % (self.user, self.hostname, " ".join(paths))
88
89
90    def _make_rsync_cmd(self, sources, dest, delete_dest, preserve_symlinks):
91        """
92        Given a list of source paths and a destination path, produces the
93        appropriate rsync command for copying them. Remote paths must be
94        pre-encoded.
95        """
96        ssh_cmd = self.make_ssh_command(user=self.user, port=self.port,
97                                        opts=self.master_ssh_option,
98                                        hosts_file=self.known_hosts_file)
99        if delete_dest:
100            delete_flag = "--delete"
101        else:
102            delete_flag = ""
103        if preserve_symlinks:
104            symlink_flag = ""
105        else:
106            symlink_flag = "-L"
107        command = ("rsync %s %s --timeout=1800 --rsh='%s' -az --no-o --no-g "
108                   "%s %s")
109        return command % (symlink_flag, delete_flag, ssh_cmd,
110                          " ".join(sources), dest)
111
112
113    def _make_ssh_cmd(self, cmd):
114        """
115        Create a base ssh command string for the host which can be used
116        to run commands directly on the machine
117        """
118        base_cmd = self.make_ssh_command(user=self.user, port=self.port,
119                                         opts=self.master_ssh_option,
120                                         hosts_file=self.known_hosts_file)
121
122        return '%s %s "%s"' % (base_cmd, self.hostname, utils.sh_escape(cmd))
123
124    def _make_scp_cmd(self, sources, dest):
125        """
126        Given a list of source paths and a destination path, produces the
127        appropriate scp command for encoding it. Remote paths must be
128        pre-encoded.
129        """
130        command = ("scp -rq %s -o StrictHostKeyChecking=no "
131                   "-o UserKnownHostsFile=%s -P %d %s '%s'")
132        return command % (self.master_ssh_option, self.known_hosts_file,
133                          self.port, " ".join(sources), dest)
134
135
136    def _make_rsync_compatible_globs(self, path, is_local):
137        """
138        Given an rsync-style path, returns a list of globbed paths
139        that will hopefully provide equivalent behaviour for scp. Does not
140        support the full range of rsync pattern matching behaviour, only that
141        exposed in the get/send_file interface (trailing slashes).
142
143        The is_local param is flag indicating if the paths should be
144        interpreted as local or remote paths.
145        """
146
147        # non-trailing slash paths should just work
148        if len(path) == 0 or path[-1] != "/":
149            return [path]
150
151        # make a function to test if a pattern matches any files
152        if is_local:
153            def glob_matches_files(path, pattern):
154                return len(glob.glob(path + pattern)) > 0
155        else:
156            def glob_matches_files(path, pattern):
157                result = self.run("ls \"%s\"%s" % (utils.sh_escape(path),
158                                                   pattern),
159                                  stdout_tee=None, ignore_status=True)
160                return result.exit_status == 0
161
162        # take a set of globs that cover all files, and see which are needed
163        patterns = ["*", ".[!.]*"]
164        patterns = [p for p in patterns if glob_matches_files(path, p)]
165
166        # convert them into a set of paths suitable for the commandline
167        if is_local:
168            return ["\"%s\"%s" % (utils.sh_escape(path), pattern)
169                    for pattern in patterns]
170        else:
171            return [utils.scp_remote_escape(path) + pattern
172                    for pattern in patterns]
173
174
175    def _make_rsync_compatible_source(self, source, is_local):
176        """
177        Applies the same logic as _make_rsync_compatible_globs, but
178        applies it to an entire list of sources, producing a new list of
179        sources, properly quoted.
180        """
181        return sum((self._make_rsync_compatible_globs(path, is_local)
182                    for path in source), [])
183
184
185    def _set_umask_perms(self, dest):
186        """
187        Given a destination file/dir (recursively) set the permissions on
188        all the files and directories to the max allowed by running umask.
189        """
190
191        # now this looks strange but I haven't found a way in Python to _just_
192        # get the umask, apparently the only option is to try to set it
193        umask = os.umask(0)
194        os.umask(umask)
195
196        max_privs = 0777 & ~umask
197
198        def set_file_privs(filename):
199            """Sets mode of |filename|.  Assumes |filename| exists."""
200            file_stat = os.stat(filename)
201
202            file_privs = max_privs
203            # if the original file permissions do not have at least one
204            # executable bit then do not set it anywhere
205            if not file_stat.st_mode & 0111:
206                file_privs &= ~0111
207
208            os.chmod(filename, file_privs)
209
210        # try a bottom-up walk so changes on directory permissions won't cut
211        # our access to the files/directories inside it
212        for root, dirs, files in os.walk(dest, topdown=False):
213            # when setting the privileges we emulate the chmod "X" behaviour
214            # that sets to execute only if it is a directory or any of the
215            # owner/group/other already has execute right
216            for dirname in dirs:
217                os.chmod(os.path.join(root, dirname), max_privs)
218
219            # Filter out broken symlinks as we go.
220            for filename in filter(os.path.exists, files):
221                set_file_privs(os.path.join(root, filename))
222
223
224        # now set privs for the dest itself
225        if os.path.isdir(dest):
226            os.chmod(dest, max_privs)
227        else:
228            set_file_privs(dest)
229
230
231    def get_file(self, source, dest, delete_dest=False, preserve_perm=True,
232                 preserve_symlinks=False):
233        """
234        Copy files from the remote host to a local path.
235
236        Directories will be copied recursively.
237        If a source component is a directory with a trailing slash,
238        the content of the directory will be copied, otherwise, the
239        directory itself and its content will be copied. This
240        behavior is similar to that of the program 'rsync'.
241
242        Args:
243                source: either
244                        1) a single file or directory, as a string
245                        2) a list of one or more (possibly mixed)
246                                files or directories
247                dest: a file or a directory (if source contains a
248                        directory or more than one element, you must
249                        supply a directory dest)
250                delete_dest: if this is true, the command will also clear
251                             out any old files at dest that are not in the
252                             source
253                preserve_perm: tells get_file() to try to preserve the sources
254                               permissions on files and dirs
255                preserve_symlinks: try to preserve symlinks instead of
256                                   transforming them into files/dirs on copy
257
258        Raises:
259                AutoservRunError: the scp command failed
260        """
261        logging.debug('get_file. source: %s, dest: %s, delete_dest: %s,'
262                      'preserve_perm: %s, preserve_symlinks:%s', source, dest,
263                      delete_dest, preserve_perm, preserve_symlinks)
264        # Start a master SSH connection if necessary.
265        self.start_master_ssh()
266
267        if isinstance(source, basestring):
268            source = [source]
269        dest = os.path.abspath(dest)
270
271        # If rsync is disabled or fails, try scp.
272        try_scp = True
273        if self.use_rsync():
274            logging.debug('Using Rsync.')
275            try:
276                remote_source = self._encode_remote_paths(source)
277                local_dest = utils.sh_escape(dest)
278                rsync = self._make_rsync_cmd([remote_source], local_dest,
279                                             delete_dest, preserve_symlinks)
280                utils.run(rsync)
281                try_scp = False
282            except error.CmdError, e:
283                logging.warning("trying scp, rsync failed: %s", e)
284
285        if try_scp:
286            logging.debug('Trying scp.')
287            # scp has no equivalent to --delete, just drop the entire dest dir
288            if delete_dest and os.path.isdir(dest):
289                shutil.rmtree(dest)
290                os.mkdir(dest)
291
292            remote_source = self._make_rsync_compatible_source(source, False)
293            if remote_source:
294                # _make_rsync_compatible_source() already did the escaping
295                remote_source = self._encode_remote_paths(remote_source,
296                                                          escape=False)
297                local_dest = utils.sh_escape(dest)
298                scp = self._make_scp_cmd([remote_source], local_dest)
299                try:
300                    utils.run(scp)
301                except error.CmdError, e:
302                    logging.debug('scp failed: %s', e)
303                    raise error.AutoservRunError(e.args[0], e.args[1])
304
305        if not preserve_perm:
306            # we have no way to tell scp to not try to preserve the
307            # permissions so set them after copy instead.
308            # for rsync we could use "--no-p --chmod=ugo=rwX" but those
309            # options are only in very recent rsync versions
310            self._set_umask_perms(dest)
311
312
313    def send_file(self, source, dest, delete_dest=False,
314                  preserve_symlinks=False):
315        """
316        Copy files from a local path to the remote host.
317
318        Directories will be copied recursively.
319        If a source component is a directory with a trailing slash,
320        the content of the directory will be copied, otherwise, the
321        directory itself and its content will be copied. This
322        behavior is similar to that of the program 'rsync'.
323
324        Args:
325                source: either
326                        1) a single file or directory, as a string
327                        2) a list of one or more (possibly mixed)
328                                files or directories
329                dest: a file or a directory (if source contains a
330                        directory or more than one element, you must
331                        supply a directory dest)
332                delete_dest: if this is true, the command will also clear
333                             out any old files at dest that are not in the
334                             source
335                preserve_symlinks: controls if symlinks on the source will be
336                    copied as such on the destination or transformed into the
337                    referenced file/directory
338
339        Raises:
340                AutoservRunError: the scp command failed
341        """
342        logging.debug('send_file. source: %s, dest: %s, delete_dest: %s,'
343                      'preserve_symlinks:%s', source, dest,
344                      delete_dest, preserve_symlinks)
345        # Start a master SSH connection if necessary.
346        self.start_master_ssh()
347
348        if isinstance(source, basestring):
349            source = [source]
350        remote_dest = self._encode_remote_paths([dest])
351
352        # If rsync is disabled or fails, try scp.
353        try_scp = True
354        if self.use_rsync():
355            logging.debug('Using Rsync.')
356            try:
357                local_sources = [utils.sh_escape(path) for path in source]
358                rsync = self._make_rsync_cmd(local_sources, remote_dest,
359                                             delete_dest, preserve_symlinks)
360                utils.run(rsync)
361                try_scp = False
362            except error.CmdError, e:
363                logging.warning("trying scp, rsync failed: %s", e)
364
365        if try_scp:
366            logging.debug('Trying scp.')
367            # scp has no equivalent to --delete, just drop the entire dest dir
368            if delete_dest:
369                is_dir = self.run("ls -d %s/" % dest,
370                                  ignore_status=True).exit_status == 0
371                if is_dir:
372                    cmd = "rm -rf %s && mkdir %s"
373                    cmd %= (dest, dest)
374                    self.run(cmd)
375
376            local_sources = self._make_rsync_compatible_source(source, True)
377            if local_sources:
378                scp = self._make_scp_cmd(local_sources, remote_dest)
379                try:
380                    utils.run(scp)
381                except error.CmdError, e:
382                    logging.debug('scp failed: %s', e)
383                    raise error.AutoservRunError(e.args[0], e.args[1])
384
385
386    def ssh_ping(self, timeout=60, base_cmd='true'):
387        """
388        Pings remote host via ssh.
389
390        @param timeout: Time in seconds before giving up.
391                        Defaults to 60 seconds.
392        @param base_cmd: The base command to run with the ssh ping.
393                         Defaults to true.
394        @raise AutoservSSHTimeout: If the ssh ping times out.
395        @raise AutoservSshPermissionDeniedError: If ssh ping fails due to
396                                                 permissions.
397        @raise AutoservSshPingHostError: For other AutoservRunErrors.
398        """
399        try:
400            self.run(base_cmd, timeout=timeout, connect_timeout=timeout)
401        except error.AutoservSSHTimeout:
402            msg = "Host (ssh) verify timed out (timeout = %d)" % timeout
403            raise error.AutoservSSHTimeout(msg)
404        except error.AutoservSshPermissionDeniedError:
405            #let AutoservSshPermissionDeniedError be visible to the callers
406            raise
407        except error.AutoservRunError, e:
408            # convert the generic AutoservRunError into something more
409            # specific for this context
410            raise error.AutoservSshPingHostError(e.description + '\n' +
411                                                 repr(e.result_obj))
412
413
414    def is_up(self, timeout=60, base_cmd='true'):
415        """
416        Check if the remote host is up by ssh-ing and running a base command.
417
418        @param timeout: timeout in seconds.
419        @param base_cmd: a base command to run with ssh. The default is 'true'.
420        @returns True if the remote host is up before the timeout expires,
421                 False otherwise.
422        """
423        try:
424            self.ssh_ping(timeout=timeout, base_cmd=base_cmd)
425        except error.AutoservError:
426            return False
427        else:
428            return True
429
430
431    def wait_up(self, timeout=None):
432        """
433        Wait until the remote host is up or the timeout expires.
434
435        In fact, it will wait until an ssh connection to the remote
436        host can be established, and getty is running.
437
438        @param timeout time limit in seconds before returning even
439            if the host is not up.
440
441        @returns True if the host was found to be up before the timeout expires,
442                 False otherwise
443        """
444        if timeout:
445            current_time = int(time.time())
446            end_time = current_time + timeout
447
448        while not timeout or current_time < end_time:
449            if self.is_up(timeout=end_time - current_time):
450                try:
451                    if self.are_wait_up_processes_up():
452                        logging.debug('Host %s is now up', self.hostname)
453                        return True
454                except error.AutoservError:
455                    pass
456            time.sleep(1)
457            current_time = int(time.time())
458
459        logging.debug('Host %s is still down after waiting %d seconds',
460                      self.hostname, int(timeout + time.time() - end_time))
461        return False
462
463
464    def wait_down(self, timeout=None, warning_timer=None, old_boot_id=None):
465        """
466        Wait until the remote host is down or the timeout expires.
467
468        If old_boot_id is provided, this will wait until either the machine
469        is unpingable or self.get_boot_id() returns a value different from
470        old_boot_id. If the boot_id value has changed then the function
471        returns true under the assumption that the machine has shut down
472        and has now already come back up.
473
474        If old_boot_id is None then until the machine becomes unreachable the
475        method assumes the machine has not yet shut down.
476
477        Based on this definition, the 4 possible permutations of timeout
478        and old_boot_id are:
479        1. timeout and old_boot_id: wait timeout seconds for either the
480                                    host to become unpingable, or the boot id
481                                    to change. In the latter case we've rebooted
482                                    and in the former case we've only shutdown,
483                                    but both cases return True.
484        2. only timeout: wait timeout seconds for the host to become unpingable.
485                         If the host remains pingable throughout timeout seconds
486                         we return False.
487        3. only old_boot_id: wait forever until either the host becomes
488                             unpingable or the boot_id changes. Return true
489                             when either of those conditions are met.
490        4. not timeout, not old_boot_id: wait forever till the host becomes
491                                         unpingable.
492
493        @param timeout Time limit in seconds before returning even
494            if the host is still up.
495        @param warning_timer Time limit in seconds that will generate
496            a warning if the host is not down yet.
497        @param old_boot_id A string containing the result of self.get_boot_id()
498            prior to the host being told to shut down. Can be None if this is
499            not available.
500
501        @returns True if the host was found to be down, False otherwise
502        """
503        #TODO: there is currently no way to distinguish between knowing
504        #TODO: boot_id was unsupported and not knowing the boot_id.
505        current_time = int(time.time())
506        if timeout:
507            end_time = current_time + timeout
508
509        if warning_timer:
510            warn_time = current_time + warning_timer
511
512        if old_boot_id is not None:
513            logging.debug('Host %s pre-shutdown boot_id is %s',
514                          self.hostname, old_boot_id)
515
516        # Impose semi real-time deadline constraints, since some clients
517        # (eg: watchdog timer tests) expect strict checking of time elapsed.
518        # Each iteration of this loop is treated as though it atomically
519        # completes within current_time, this is needed because if we used
520        # inline time.time() calls instead then the following could happen:
521        #
522        # while not timeout or time.time() < end_time:      [23 < 30]
523        #    some code.                                     [takes 10 secs]
524        #    try:
525        #        new_boot_id = self.get_boot_id(timeout=end_time - time.time())
526        #                                                   [30 - 33]
527        # The last step will lead to a return True, when in fact the machine
528        # went down at 32 seconds (>30). Hence we need to pass get_boot_id
529        # the same time that allowed us into that iteration of the loop.
530        while not timeout or current_time < end_time:
531            try:
532                new_boot_id = self.get_boot_id(timeout=end_time-current_time)
533            except error.AutoservError:
534                logging.debug('Host %s is now unreachable over ssh, is down',
535                              self.hostname)
536                return True
537            else:
538                # if the machine is up but the boot_id value has changed from
539                # old boot id, then we can assume the machine has gone down
540                # and then already come back up
541                if old_boot_id is not None and old_boot_id != new_boot_id:
542                    logging.debug('Host %s now has boot_id %s and so must '
543                                  'have rebooted', self.hostname, new_boot_id)
544                    return True
545
546            if warning_timer and current_time > warn_time:
547                self.record("INFO", None, "shutdown",
548                            "Shutdown took longer than %ds" % warning_timer)
549                # Print the warning only once.
550                warning_timer = None
551                # If a machine is stuck switching runlevels
552                # This may cause the machine to reboot.
553                self.run('kill -HUP 1', ignore_status=True)
554
555            time.sleep(1)
556            current_time = int(time.time())
557
558        return False
559
560
561    # tunable constants for the verify & repair code
562    AUTOTEST_GB_DISKSPACE_REQUIRED = get_value("SERVER",
563                                               "gb_diskspace_required",
564                                               type=float,
565                                               default=20.0)
566
567
568    def verify_connectivity(self):
569        super(AbstractSSHHost, self).verify_connectivity()
570
571        logging.info('Pinging host ' + self.hostname)
572        self.ssh_ping()
573        logging.info("Host (ssh) %s is alive", self.hostname)
574
575        if self.is_shutting_down():
576            raise error.AutoservHostIsShuttingDownError("Host is shutting down")
577
578
579    def verify_software(self):
580        super(AbstractSSHHost, self).verify_software()
581        try:
582            self.check_diskspace(autotest.Autotest.get_install_dir(self),
583                                 self.AUTOTEST_GB_DISKSPACE_REQUIRED)
584        except error.AutoservHostError:
585            raise           # only want to raise if it's a space issue
586        except autotest.AutodirNotFoundError:
587            # autotest dir may not exist, etc. ignore
588            logging.debug('autodir space check exception, this is probably '
589                          'safe to ignore\n' + traceback.format_exc())
590
591
592    def close(self):
593        super(AbstractSSHHost, self).close()
594        self._cleanup_master_ssh()
595        os.remove(self.known_hosts_file)
596
597
598    def _cleanup_master_ssh(self):
599        """
600        Release all resources (process, temporary directory) used by an active
601        master SSH connection.
602        """
603        # If a master SSH connection is running, kill it.
604        if self.master_ssh_job is not None:
605            logging.debug('Nuking master_ssh_job.')
606            utils.nuke_subprocess(self.master_ssh_job.sp)
607            self.master_ssh_job = None
608
609        # Remove the temporary directory for the master SSH socket.
610        if self.master_ssh_tempdir is not None:
611            logging.debug('Cleaning master_ssh_tempdir.')
612            self.master_ssh_tempdir.clean()
613            self.master_ssh_tempdir = None
614            self.master_ssh_option = ''
615
616
617    def start_master_ssh(self, timeout=5):
618        """
619        Called whenever a slave SSH connection needs to be initiated (e.g., by
620        run, rsync, scp). If master SSH support is enabled and a master SSH
621        connection is not active already, start a new one in the background.
622        Also, cleanup any zombie master SSH connections (e.g., dead due to
623        reboot).
624
625        timeout: timeout in seconds (default 5) to wait for master ssh
626                 connection to be established. If timeout is reached, a
627                 warning message is logged, but no other action is taken.
628        """
629        if not enable_master_ssh:
630            return
631
632        # If a previously started master SSH connection is not running
633        # anymore, it needs to be cleaned up and then restarted.
634        if self.master_ssh_job is not None:
635            socket_path = os.path.join(self.master_ssh_tempdir.name, 'socket')
636            if (not os.path.exists(socket_path) or
637                self.master_ssh_job.sp.poll() is not None):
638                if self.master_ssh_job.sp.poll() is None:
639                    logging.warning('Master ssh connection socket file '
640                                    'was missing while its subprocess was '
641                                    'still running.')
642                    if os.path.exists(self.master_ssh_tempdir.name):
643                        logging.warning('However, the socket file temporary '
644                                        'directory still exists.')
645
646                    logging.warning('Info on defunct master ssh ps below.')
647                    master_pid = str(self.master_ssh_job.sp.pid)
648                    ps_output = utils.run(['ps', '-Fww', master_pid],
649                                          ignore_status=True).stdout
650                    logging.warning('Master ssh connection ps info: %s',
651                                    ps_output)
652                    lsof_output = utils.run(['lsof', '-p', master_pid],
653                                            ignore_status=True).stdout
654                    logging.warning('Master ssh connection lsof info: %s',
655                                    lsof_output)
656
657                logging.info("Master ssh connection to %s is down.",
658                             self.hostname)
659                self._cleanup_master_ssh()
660
661        # Start a new master SSH connection.
662        if self.master_ssh_job is None:
663            # Create a shared socket in a temp location.
664            self.master_ssh_tempdir = autotemp.tempdir(unique_id='ssh-master')
665            self.master_ssh_option = ("-o ControlPath=%s/socket" %
666                                      self.master_ssh_tempdir.name)
667
668            # Start the master SSH connection in the background.
669            master_cmd = self.ssh_command(options="-N -o ControlMaster=yes")
670            logging.debug("System load: %s",  utils.run(['uptime']).stdout)
671            logging.info("Starting master ssh connection '%s'", master_cmd)
672            self.master_ssh_job = utils.BgJob(master_cmd,
673                                              nickname='master-ssh',
674                                              no_pipes=True)
675            # To prevent a race between the the master ssh connection startup
676            # and its first attempted use, wait for socket file to exist before
677            # returning.
678            end_time = time.time() + timeout
679            socket_file_path = os.path.join(self.master_ssh_tempdir.name,
680                                            'socket')
681            while time.time() < end_time:
682                if os.path.exists(socket_file_path):
683                    break
684                time.sleep(.2)
685            else:
686                logging.warning('Timed out waiting for master-ssh connection '
687                             'to be established.')
688
689
690    def clear_known_hosts(self):
691        """Clears out the temporary ssh known_hosts file.
692
693        This is useful if the test SSHes to the machine, then reinstalls it,
694        then SSHes to it again.  It can be called after the reinstall to
695        reduce the spam in the logs.
696        """
697        logging.info("Clearing known hosts for host '%s', file '%s'.",
698                     self.hostname, self.known_hosts_file)
699        # Clear out the file by opening it for writing and then closing.
700        fh = open(self.known_hosts_file, "w")
701        fh.close()
702
703
704    def collect_logs(self, remote_src_dir, local_dest_dir, ignore_errors=True):
705        """Copy log directories from a host to a local directory.
706
707        @param remote_src_dir: A destination directory on the host.
708        @param local_dest_dir: A path to a local destination directory.
709            If it doesn't exist it will be created.
710        @param ignore_errors: If True, ignore exceptions.
711
712        @raises OSError: If there were problems creating the local_dest_dir and
713            ignore_errors is False.
714        @raises AutoservRunError, AutotestRunError: If something goes wrong
715            while copying the directories and ignore_errors is False.
716        """
717        locally_created_dest = False
718        if (not os.path.exists(local_dest_dir)
719                or not os.path.isdir(local_dest_dir)):
720            try:
721                os.makedirs(local_dest_dir)
722                locally_created_dest = True
723            except OSError as e:
724                logging.warning('Unable to collect logs from host '
725                                '%s: %s', self.hostname, e)
726                if not ignore_errors:
727                    raise
728                return
729        try:
730            self.get_file(
731                    remote_src_dir, local_dest_dir, preserve_symlinks=True)
732        except (error.AutotestRunError, error.AutoservRunError,
733                error.AutoservSSHTimeout) as e:
734            logging.warning('Collection of %s to local dir %s from host %s '
735                            'failed: %s', remote_src_dir, local_dest_dir,
736                            self.hostname, e)
737            if locally_created_dest:
738                shutil.rmtree(local_dest_dir, ignore_errors=ignore_errors)
739            if not ignore_errors:
740                raise
741
742
743