From f0224ce93ae9ad420e23612fe6e6707a821f9cab Mon Sep 17 00:00:00 2001 From: Kotresh HR Date: Mon, 23 Feb 2015 14:46:48 +0530 Subject: feature/geo-rep: Active Passive Switching logic flock CURRENT DESIGN AND ITS LIMITATIONS: ----------------------------------- Geo-replication syncs changes across geography using changelogs captured by changelog translator. Changelog translator sits on server side just above posix translator. Hence, in distributed replicated setup, both replica pairs collect changelogs w.r.t their bricks. Geo-replication syncs the changes using only one brick among the replica pair at a time, calling it as "ACTIVE" and other non syncing brick as "PASSIVE". Let's consider below example of distributed replicated setup where NODE-1 as b1 and its replicated brick b1r is in NODE-2 NODE-1 NODE-2 b1 b1r At the beginning, geo-replication chooses to sync changes from NODE-1:b1 and NODE-2:b1r will be "PASSIVE". The logic depends on virtual getxattr 'trusted.glusterfs.node-uuid' which always returns first up subvolume i.e., NODE-1. When NODE-1 goes down, the above xattr returns NODE-2 and that is made 'ACTIVE'. But when NODE-1 comes back again, the above xattr returns NODE-1 and it is made 'ACTIVE' again. So for a brief interval of time, if NODE-2 had not finished processing the changelog, both NODE-2 and NODE-1 will be ACTIVE causing rename race as mentioned in the bug. SOLUTION: --------- 1. Have a shared replicated storage, a glusterfs management volume specific to geo-replication. 2. Geo-rep creates a file per replica set on management volume. 3. fcntl lock on the above said file is used for synchronization between geo-rep workers belonging to same replica set. 4. If management volume is not configured, geo-replication will back to previous logic of using first up sub volume. Each worker tries to lock the file on shared storage, who ever wins will be ACTIVE. With this, we are able to solve the problem but there is an issue when the shared replicated storage goes down (when all replicas goes down). In that case, the lock state is lost. So AFR needs to rebuild the lock state after brick comes up. NOTE: ----- This patch brings in the, pre-requisite step of setting up management volume for geo-replication during creation. 1. Create mgmt-vol for geo-replicatoin and start it. Management volume should be part of master cluster and recommended to be three way replicated volume having each brick in different nodes for availability. 2. Create geo-rep session. 3. Configure mgmt-vol created with geo-replication session as follows. gluster vol geo-rep slavenode:: config meta_volume \ 4. Start geo-rep session. Backward Compatiability: ----------------------- If management volume is not configured, it falls back to previous logic of using node-uuid virtual xattr. But it is not recommended. Change-Id: I7319d2289516f534b69edd00c9d0db5a3725661a BUG: 1196632 Signed-off-by: Kotresh HR Reviewed-on: http://review.gluster.org/9759 Tested-by: Gluster Build System Reviewed-by: Vijay Bellur --- geo-replication/syncdaemon/gsyncd.py | 4 +++ geo-replication/syncdaemon/master.py | 39 ++++++++++++++++++++++++--- geo-replication/syncdaemon/monitor.py | 45 +++++++++++++++++++++++++++++++- geo-replication/syncdaemon/syncdutils.py | 4 +++ 4 files changed, 88 insertions(+), 4 deletions(-) (limited to 'geo-replication/syncdaemon') diff --git a/geo-replication/syncdaemon/gsyncd.py b/geo-replication/syncdaemon/gsyncd.py index b38f19d33f7..1542810bcd7 100644 --- a/geo-replication/syncdaemon/gsyncd.py +++ b/geo-replication/syncdaemon/gsyncd.py @@ -255,6 +255,8 @@ def main_i(): type=int, default=1) op.add_option('--changelog-archive-format', metavar='N', type=str, default="%Y%m") + op.add_option('--meta-volume', metavar='N', + type=str, default="") op.add_option( '--turns', metavar='N', type=int, default=0, help=SUPPRESS_HELP) op.add_option('--allow-network', metavar='IPS', default='') @@ -297,6 +299,8 @@ def main_i(): op.add_option('--feedback-fd', dest='feedback_fd', type=int, help=SUPPRESS_HELP, action='callback', callback=store_local) op.add_option('--rpc-fd', dest='rpc_fd', type=str, help=SUPPRESS_HELP) + op.add_option('--subvol-num', dest='subvol_num', type=int, + help=SUPPRESS_HELP) op.add_option('--listen', dest='listen', help=SUPPRESS_HELP, action='callback', callback=store_local_curry(True)) op.add_option('-N', '--no-daemon', dest="go_daemon", diff --git a/geo-replication/syncdaemon/master.py b/geo-replication/syncdaemon/master.py index 51c26c76116..dfe65fe6709 100644 --- a/geo-replication/syncdaemon/master.py +++ b/geo-replication/syncdaemon/master.py @@ -14,11 +14,12 @@ import time import stat import json import logging +import fcntl import socket import string import errno import tarfile -from errno import ENOENT, ENODATA, EPIPE, EEXIST +from errno import ENOENT, ENODATA, EPIPE, EEXIST, EACCES, EAGAIN from threading import Condition, Lock from datetime import datetime from gconf import gconf @@ -452,8 +453,40 @@ class GMasterCommon(object): t = Thread(target=keep_alive) t.start() - def should_crawl(cls): - return gconf.glusterd_uuid in cls.master.server.node_uuid() + def mgmt_lock(self): + """Take management volume lock """ + bname = str(gconf.volume_id) + "_subvol_" + str(gconf.subvol_num) \ + + ".lock" + path = os.path.join(gconf.working_dir, gconf.meta_volume, bname) + logging.debug("lock_file_path: %s" % path) + fd = os.open(path, os.O_CREAT | os.O_RDWR) + try: + fcntl.lockf(fd, fcntl.LOCK_EX | fcntl.LOCK_NB) + except: + ex = sys.exc_info()[1] + os.close(fd) + if isinstance(ex, IOError) and ex.errno in (EACCES, EAGAIN): + # cannot grab, it's taken + logging.debug("Lock held by someother worker process") + return False + raise + logging.debug("Got the lock") + return True + + + def should_crawl(self): + if not gconf.meta_volume: + return gconf.glusterd_uuid in self.master.server.node_uuid() + + mgmt_mnt = os.path.join(gconf.working_dir, gconf.meta_volume) + if not os.path.ismount(mgmt_mnt): + po = Popen(["mount", "-t", "glusterfs", "localhost:%s" + % gconf.meta_volume, mgmt_mnt], stdout=PIPE, + stderr=PIPE) + po.wait() + po.terminate_geterr() + return self.mgmt_lock() + def register(self): self.register() diff --git a/geo-replication/syncdaemon/monitor.py b/geo-replication/syncdaemon/monitor.py index 5e0698b8c46..e50893c793f 100644 --- a/geo-replication/syncdaemon/monitor.py +++ b/geo-replication/syncdaemon/monitor.py @@ -18,6 +18,7 @@ import xml.etree.ElementTree as XET from subprocess import PIPE from resource import Popen, FILE, GLUSTER, SSH from threading import Lock +from errno import EEXIST import re import random from gconf import gconf @@ -29,6 +30,16 @@ from syncdutils import escape, Thread, finalize, memoize ParseError = XET.ParseError if hasattr(XET, 'ParseError') else SyntaxError +def get_subvol_num(brick_idx, replica_count, disperse_count): + subvol_size = disperse_count if disperse_count > 0 else replica_count + cnt = int((brick_idx + 1) / subvol_size) + rem = (brick_idx + 1) % subvol_size + if rem > 0: + return cnt + 1 + else: + return cnt + + def get_slave_bricks_status(host, vol): po = Popen(['gluster', '--xml', '--remote-host=' + host, 'volume', 'status', vol, "detail"], @@ -104,6 +115,15 @@ class Volinfo(object): self.volume, self.host) return ids[0].text + @property + @memoize + def replica_count(self): + return int(self.get('replicaCount')[0].text) + + @property + @memoize + def disperse_count(self): + return int(self.get('disperseCount')[0].text) class Monitor(object): @@ -154,6 +174,7 @@ class Monitor(object): # give a chance to graceful exit os.kill(-os.getpid(), signal.SIGTERM) + def monitor(self, w, argv, cpids, agents, slave_vol, slave_host): """the monitor loop @@ -247,6 +268,7 @@ class Monitor(object): '--rpc-fd', ','.join([str(rw), str(ww), str(ra), str(wa)]), + '--subvol-num', str(w[2]), '--resource-remote', remote_host]) @@ -374,7 +396,8 @@ def distribute(*resources): else: slaves = slavevols - workerspex = [(brick['dir'], slaves[idx % len(slaves)]) + workerspex = [(brick['dir'], slaves[idx % len(slaves)], + get_subvol_num(idx, mvol.replica_count, mvol.disperse_count)) for idx, brick in enumerate(mvol.bricks) if is_host_local(brick['host'])] logging.info('worker specs: ' + repr(workerspex)) @@ -382,6 +405,26 @@ def distribute(*resources): def monitor(*resources): + # Mount geo-rep management volume + if gconf.meta_volume: + mgmt_mnt = os.path.join(gconf.working_dir, gconf.meta_volume) + if not os.path.exists(mgmt_mnt): + try: + os.makedirs(mgmt_mnt) + except OSError: + ex = sys.exc_info()[1] + if ex.errno == EEXIST: + pass + else: + raise + + if not os.path.ismount(mgmt_mnt): + po = Popen(["mount", "-t", "glusterfs", "localhost:%s" + % gconf.meta_volume, mgmt_mnt], stdout=PIPE, + stderr=PIPE) + po.wait() + po.terminate_geterr() + # Check if gsyncd restarted in pause state. If # yes, send SIGSTOP to negative of monitor pid # to go back to pause state. diff --git a/geo-replication/syncdaemon/syncdutils.py b/geo-replication/syncdaemon/syncdutils.py index 6af957ddb4a..5037004d9a5 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 +from subprocess import Popen, PIPE from threading import Lock, Thread as baseThread from errno import EACCES, EAGAIN, EPIPE, ENOTCONN, ECONNABORTED from errno import EINTR, ENOENT, EPERM, ESTALE, errorcode @@ -215,6 +216,9 @@ def finalize(*a, **kw): except: if sys.exc_info()[0] == OSError: pass + + # TODO: Clean up mgmt volume mount point only monitor dies + if gconf.log_exit: logging.info("exiting.") sys.stdout.flush() -- cgit