From fa2086eaa9545b472acd3fcc07be776d9f2cb38a Mon Sep 17 00:00:00 2001

From: Kotresh HR <khiremat@redhat.com>

Date: Thu, 15 Feb 2018 01:46:29 -0500

Subject: [PATCH 646/646] geo-rep: Remove lazy umount and use mount namespaces

Lazy umounting the master volume by worker causes

issues with rsync's usage of getcwd. Henc removing

the lazy umount and using private mount namespace

for the same. On the slave, the lazy umount is

retained as we can't use private namespace in non

root geo-rep setup because gsyncd is spawned as

non privileged user.

Backport of https://review.gluster.org/#/c/19544/

Change-Id: I851e8dc2b8523dc5668a97e87ef619ab70471dfd

BUG: 1547931

Signed-off-by: Kotresh HR <khiremat@redhat.com>

Reviewed-on: https://code.engineering.redhat.com/gerrit/131128

Tested-by: RHGS Build Bot <nigelb@redhat.com>

Reviewed-by: Aravinda Vishwanathapura Krishna Murthy <avishwan@redhat.com>

Reviewed-by: Atin Mukherjee <amukherj@redhat.com>

---

 geo-replication/syncdaemon/gconf.py      |  3 +++

 geo-replication/syncdaemon/gsyncd.py     | 14 +++++++-----

 geo-replication/syncdaemon/monitor.py    | 38 ++++++++++++++++++++------------

 geo-replication/syncdaemon/resource.py   | 16 ++++++++++++--

 geo-replication/syncdaemon/syncdutils.py | 18 ++++++++++++++-

 glusterfs.spec.in                        |  4 ++++

 6 files changed, 70 insertions(+), 23 deletions(-)

diff --git a/geo-replication/syncdaemon/gconf.py b/geo-replication/syncdaemon/gconf.py

index 97395b4..2280f44 100644

--- a/geo-replication/syncdaemon/gconf.py

+++ b/geo-replication/syncdaemon/gconf.py

@@ -28,5 +28,8 @@ class GConf(object):

     active_earlier = False

     passive_earlier = False

     mgmt_lock_fd = None

+    mountbroker = False

+    mount_point = None

+    mbr_umount_cmd = []

 gconf = GConf()

diff --git a/geo-replication/syncdaemon/gsyncd.py b/geo-replication/syncdaemon/gsyncd.py

index f9471e4..96256cf 100644

--- a/geo-replication/syncdaemon/gsyncd.py

+++ b/geo-replication/syncdaemon/gsyncd.py

@@ -269,6 +269,8 @@ def main_i():

                   type=str, action='callback', callback=store_abs)

     op.add_option('--georep-session-working-dir', metavar='STATF',

                   type=str, action='callback', callback=store_abs)

+    op.add_option('--access-mount', default=False, action='store_true')

+    op.add_option('--slave-access-mount', default=False, action='store_true')

     op.add_option('--ignore-deletes', default=False, action='store_true')

     op.add_option('--isolated-slave', default=False, action='store_true')

     op.add_option('--use-rsync-xattrs', default=False, action='store_true')

@@ -414,7 +416,7 @@ def main_i():

                     o.get_opt_string() not in ('--version', '--help'))]

     remote_tunables = ['listen', 'go_daemon', 'timeout',

                        'session_owner', 'config_file', 'use_rsync_xattrs',

-                       'local_id', 'local_node']

+                       'local_id', 'local_node', 'slave_access_mount']

     rq_remote_tunables = {'listen': True}

     # precedence for sources of values: 1) commandline, 2) cfg file, 3)

@@ -748,15 +750,15 @@ def main_i():

     else:

         log_file = gconf.log_file

     if be_monitor:

-        label = 'monitor'

+        gconf.label = 'monitor'

     elif be_agent:

-        label = gconf.local_path

+        gconf.label = gconf.local_path

     elif remote:

         # master

-        label = gconf.local_path

+        gconf.label = gconf.local_path

     else:

-        label = 'slave'

-    startup(go_daemon=go_daemon, log_file=log_file, label=label)

+        gconf.label = 'slave'

+    startup(go_daemon=go_daemon, log_file=log_file, label=gconf.label)

     resource.Popen.init_errhandler()

     if be_agent:

diff --git a/geo-replication/syncdaemon/monitor.py b/geo-replication/syncdaemon/monitor.py

index dc0211e..087a202 100644

--- a/geo-replication/syncdaemon/monitor.py

+++ b/geo-replication/syncdaemon/monitor.py

@@ -24,7 +24,7 @@ import random

 from gconf import gconf

 from syncdutils import select, waitpid, errno_wrap

 from syncdutils import set_term_handler, is_host_local, GsyncdError

-from syncdutils import escape, Thread, finalize, memoize

+from syncdutils import escape, Thread, finalize, memoize, boolify

 from syncdutils import gf_event, EVENT_GEOREP_FAULTY

 from gsyncdstatus import GeorepStatus, set_monitor_status

@@ -301,19 +301,29 @@ class Monitor(object):

                 os.close(pr)

                 os.close(ra)

                 os.close(wa)

-                os.execv(sys.executable, argv + ['--feedback-fd', str(pw),

-                                                 '--local-path', w[0]['dir'],

-                                                 '--local-node', w[0]['host'],

-                                                 '--local-node-id',

-                                                 w[0]['uuid'],

-                                                 '--local-id',

-                                                 '.' + escape(w[0]['dir']),

-                                                 '--rpc-fd',

-                                                 ','.join([str(rw), str(ww),

-                                                           str(ra), str(wa)]),

-                                                 '--subvol-num', str(w[2])] +

-                         (['--is-hottier'] if w[3] else []) +

-                         ['--resource-remote', remote_host])

+                args_to_worker = argv + ['--feedback-fd', str(pw),

+                                         '--local-path', w[0]['dir'],

+                                         '--local-node', w[0]['host'],

+                                         '--local-node-id',

+                                         w[0]['uuid'],

+                                         '--local-id',

+                                         '.' + escape(w[0]['dir']),

+                                         '--rpc-fd',

+                                         ','.join([str(rw), str(ww),

+                                         str(ra), str(wa)]),

+                                         '--subvol-num', str(w[2])]

+

+                if w[3]:

+                    args_to_worker.append('--is-hottier')

+                args_to_worker += ['--resource-remote', remote_host]

+

+                access_mount = boolify(gconf.access_mount)

+                if access_mount:

+                    os.execv(sys.executable, args_to_worker)

+                else:

+                    unshare_cmd = ['unshare', '-m', '--propagation', 'private']

+                    cmd = unshare_cmd + args_to_worker

+                    os.execvp("unshare", cmd)

             cpids.add(cpid)

             agents.add(apid)

diff --git a/geo-replication/syncdaemon/resource.py b/geo-replication/syncdaemon/resource.py

index 943e3ec..39d537b 100644

--- a/geo-replication/syncdaemon/resource.py

+++ b/geo-replication/syncdaemon/resource.py

@@ -989,6 +989,8 @@ class SlaveRemote(object):

             extra_opts += ['--local-node', ln]

         if boolify(gconf.use_rsync_xattrs):

             extra_opts.append('--use-rsync-xattrs')

+        if boolify(gconf.slave_access_mount):

+            extra_opts.append('--slave-access-mount')

         po = Popen(rargs + gconf.remote_gsyncd.split() + extra_opts +

                    ['-N', '--listen', '--timeout', str(gconf.timeout),

                     slave],

@@ -1258,6 +1260,7 @@ class GLUSTER(AbstractUrl, SlaveLocal, SlaveRemote):

         def __init__(self, params):

             self.params = params

             self.mntpt = None

+            self.umount_cmd = []

         @classmethod

         def get_glusterprog(cls):

@@ -1348,13 +1351,16 @@ class GLUSTER(AbstractUrl, SlaveLocal, SlaveRemote):

                         assert(mntdata[-1] == '\0')

                         mntpt = mntdata[:-1]

                         assert(mntpt)

-                        if mounted:

+                        if mounted and gconf.label == 'slave' \

+                           and not boolify(gconf.slave_access_mount):

                             po = self.umount_l(mntpt)

                             po.terminate_geterr(fail_on_err=False)

                             if po.returncode != 0:

                                 po.errlog()

                                 rv = po.returncode

-                        self.cleanup_mntpt(mntpt)

+                        if gconf.label == 'slave' \

+                           and not boolify(gconf.slave_access_mount):

+                            self.cleanup_mntpt(mntpt)

                 except:

                     logging.exception('mount cleanup failure:')

                     rv = 200

@@ -1374,6 +1380,7 @@ class GLUSTER(AbstractUrl, SlaveLocal, SlaveRemote):

         def make_mount_argv(self):

             self.mntpt = tempfile.mkdtemp(prefix='gsyncd-aux-mount-')

+            gconf.mount_point = self.mntpt

             return [self.get_glusterprog()] + \

                 ['--' + p for p in self.params] + [self.mntpt]

@@ -1405,6 +1412,11 @@ class GLUSTER(AbstractUrl, SlaveLocal, SlaveRemote):

         def handle_mounter(self, po):

             self.mntpt = po.stdout.readline()[:-1]

+            gconf.mount_point = self.mntpt

+            gconf.mountbroker = True

+            self.umount_cmd = self.make_cli_argv() + ['umount']

+            gconf.mbr_umount_cmd = self.umount_cmd

+

             po.stdout.close()

             sup(self, po)

             if po.returncode != 0:

diff --git a/geo-replication/syncdaemon/syncdutils.py b/geo-replication/syncdaemon/syncdutils.py

index a22289e..8dc6c96 100644

--- a/geo-replication/syncdaemon/syncdutils.py

+++ b/geo-replication/syncdaemon/syncdutils.py

@@ -16,6 +16,7 @@ import fcntl

 import shutil

 import logging

 import socket

+import subprocess

 from threading import Lock, Thread as baseThread

 from errno import EACCES, EAGAIN, EPIPE, ENOTCONN, ECONNABORTED

 from errno import EINTR, ENOENT, EPERM, ESTALE, EBUSY, errorcode

@@ -188,7 +189,6 @@ def grabpidfile(fname=None, setpid=True):

 final_lock = Lock()

-

 def finalize(*a, **kw):

     """all those messy final steps we go trough upon termination

@@ -233,6 +233,22 @@ def finalize(*a, **kw):

             if sys.exc_info()[0] == OSError:

                 pass

+    """ Unmount if not done """

+    if gconf.mount_point:

+        if gconf.mountbroker:

+            umount_cmd = gconf.mbr_umount_cmd + [gconf.mount_point, 'lazy']

+        else:

+            umount_cmd = ['umount', '-l', gconf.mount_point]

+        p0 = subprocess.Popen(umount_cmd, stderr=subprocess.PIPE)

+        _, errdata = p0.communicate()

+        if p0.returncode == 0:

+            try:

+                os.rmdir(gconf.mount_point)

+            except OSError:

+                pass

+        else:

+            pass

+

     if gconf.log_exit:

         logging.info("exiting.")

     sys.stdout.flush()

diff --git a/glusterfs.spec.in b/glusterfs.spec.in

index fc9125b..b7bfcac 100644

--- a/glusterfs.spec.in

+++ b/glusterfs.spec.in

@@ -439,6 +439,7 @@ Requires:         %{name}%{?_isa} = %{version}-%{release}

 Requires:         %{name}-server%{?_isa} = %{version}-%{release}

 Requires:         python python-ctypes

 Requires:         rsync

+Requires:         util-linux

 %description geo-replication

 GlusterFS is a distributed file-system capable of scaling to several

@@ -2067,6 +2068,9 @@ fi

 %endif

 %changelog

+* Thu Feb 22 2018 Kotresh HR <khiremat@redhat.com>

+- Added util-linux as dependency to georeplication rpm (#1544382)

+

 * Wed Dec 20 2017 Milind Changire <mchangir@redhat.com>

 - Remove ExclusiveArch directive to help building on non-x86_64 arches (#1527772)

-- 

1.8.3.1