mirror of
https://github.com/xcat2/confluent.git
synced 2026-09-05 04:27:56 +00:00
Jarrod Johnson
687b8e47e4
Directories are left as 'boring' confluent directories to enable staging. Then the ownership/permssions on directories are fixed up. Then after completion, make sure ownership is back to boring before asking rmtree.
402 lines
16 KiB
Python
402 lines
16 KiB
Python
# vim: tabstop=4 shiftwidth=4 softtabstop=4
|
|
|
|
# Copyright 2021 Lenovo
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
import asyncio
|
|
import glob
|
|
import os
|
|
import subprocess
|
|
import shutil
|
|
import tempfile
|
|
import confluent.sshutil as sshutil
|
|
import confluent.util as util
|
|
import confluent.noderange as noderange
|
|
import confluent.tasks as tasks
|
|
import pwd
|
|
import grp
|
|
import sys
|
|
|
|
def mkdirp(path):
|
|
try:
|
|
os.makedirs(path)
|
|
except OSError as e:
|
|
if e.errno != 17:
|
|
raise
|
|
|
|
def get_entries(filename):
|
|
secname = 'REPLACE:'
|
|
filename = filename.strip()
|
|
if filename[-1] == '>':
|
|
filename = filename[:-1]
|
|
with open(filename, 'r') as slfile:
|
|
slist = slfile.read()
|
|
entries = slist.split('\n')
|
|
for ent in entries:
|
|
ent = ent.split('#', 1)[0].strip()
|
|
if not ent:
|
|
continue
|
|
if ent in ('APPENDONCE:', 'MERGE:', 'REPLACE:'):
|
|
secname = ent
|
|
if ent[0] == '<':
|
|
subfilename = ent[1:]
|
|
if subfilename[-1] == '>':
|
|
subfilename = subfilename[:-1]
|
|
if subfilename[0] != '/':
|
|
subfilename = os.path.join(os.path.dirname(filename), subfilename)
|
|
for subent in get_entries(subfilename):
|
|
yield subent
|
|
yield secname
|
|
else:
|
|
yield ent
|
|
|
|
class SyncList(object):
|
|
def __init__(self, filename, nodename, cfg):
|
|
self.replacemap = {}
|
|
self.appendmap = {}
|
|
self.appendoncemap = {}
|
|
self.mergemap = {}
|
|
self.optmap = {}
|
|
entries = get_entries(filename)
|
|
currmap = self.replacemap
|
|
for ent in entries:
|
|
try:
|
|
cmtidx = ent.index('#')
|
|
ent = ent[:cmtidx]
|
|
except ValueError:
|
|
pass
|
|
for special in '$%^&|{}':
|
|
if special in ent:
|
|
raise Exception(
|
|
'Special character "{}" reserved for future use'.format(special))
|
|
ent = ent.strip()
|
|
if not ent:
|
|
continue
|
|
if ent[-1] == ':':
|
|
if ent == 'MERGE:':
|
|
currmap = self.mergemap
|
|
elif ent == 'APPENDONCE:':
|
|
currmap = self.appendoncemap
|
|
elif ent == 'REPLACE:':
|
|
currmap = self.replacemap
|
|
else:
|
|
raise Exception(
|
|
'Section "{}" is not currently supported in syncfiles'.format(ent[:-1]))
|
|
continue
|
|
if '->' in ent:
|
|
k, v = ent.split('->')
|
|
k = k.strip()
|
|
v = v.strip()
|
|
if ':' in v:
|
|
nr, v = v.split(':', 1)
|
|
try:
|
|
for candidate in noderange.NodeRange(nr, cfg).nodes:
|
|
if candidate == nodename:
|
|
break
|
|
else:
|
|
continue
|
|
except Exception as e:
|
|
raise Exception('Error on syncfile line "{}": {}'.format(ent, str(e)))
|
|
optparts = v.split()
|
|
v = optparts[0]
|
|
optparts = optparts[1:]
|
|
else:
|
|
kparts = []
|
|
optparts = []
|
|
currparts = kparts
|
|
for part in ent.split():
|
|
if part[0] == '(':
|
|
currparts = optparts
|
|
currparts.append(part)
|
|
k = ' '.join(kparts)
|
|
v = None
|
|
entopts = {}
|
|
if optparts:
|
|
if optparts[0][0] != '(' or optparts[-1][-1] != ')':
|
|
raise Exception("Unsupported syntax in syncfile: " + ent)
|
|
opts = ','.join(optparts)
|
|
opts = opts[1:-1]
|
|
for opt in opts.split(','):
|
|
optname, optval = opt.split('=')
|
|
if optname == 'owner':
|
|
try:
|
|
uid = pwd.getpwnam(optval).pw_uid
|
|
except KeyError:
|
|
try:
|
|
uid = int(optval)
|
|
optval = None
|
|
except ValueError:
|
|
uid = None
|
|
if optval:
|
|
optval = {'name': optval}
|
|
else:
|
|
optval = {}
|
|
if uid is not None:
|
|
optval['id'] = uid
|
|
elif optname == 'group':
|
|
try:
|
|
gid = grp.getgrnam(optval).gr_gid
|
|
except KeyError:
|
|
try:
|
|
gid = int(optval)
|
|
optval = None
|
|
except ValueError:
|
|
gid = None
|
|
if optval:
|
|
optval = {'name': optval}
|
|
else:
|
|
optval = {}
|
|
if gid is not None:
|
|
optval['id'] = gid
|
|
if optval:
|
|
entopts[optname] = optval
|
|
currmap[k] = v
|
|
targ = v if v else k
|
|
for f in targ.split():
|
|
self.optmap[f] = entopts
|
|
|
|
|
|
async def sync_list_to_node(sl, node, suffixes, peerip=None):
|
|
targdir = tempfile.mkdtemp('.syncto{}'.format(node))
|
|
output = ''
|
|
correctionlist = []
|
|
try:
|
|
for ent in sl.replacemap:
|
|
stage_ent(sl.replacemap, ent, targdir, correctionlist=correctionlist)
|
|
if 'append' in suffixes:
|
|
while suffixes['append'] and suffixes['append'][0] == '/':
|
|
suffixes['append'] = suffixes['append'][1:]
|
|
for ent in sl.appendmap:
|
|
stage_ent(sl.appendmap, ent,
|
|
os.path.join(targdir, suffixes['append']), correctionlist=correctionlist)
|
|
if 'merge' in suffixes:
|
|
while suffixes['merge'] and suffixes['merge'][0] == '/':
|
|
suffixes['merge'] = suffixes['merge'][1:]
|
|
for ent in sl.mergemap:
|
|
stage_ent(sl.mergemap, ent,
|
|
os.path.join(targdir, suffixes['merge']), True, correctionlist=correctionlist)
|
|
if 'appendonce' in suffixes:
|
|
while suffixes['appendonce'] and suffixes['appendonce'][0] == '/':
|
|
suffixes['appendonce'] = suffixes['appendonce'][1:]
|
|
for ent in sl.appendoncemap:
|
|
stage_ent(sl.appendoncemap, ent,
|
|
os.path.join(targdir, suffixes['appendonce']), True, correctionlist=correctionlist)
|
|
for correction in correctionlist:
|
|
source, destination, srcstat = correction
|
|
shutil.copystat(source, destination)
|
|
try:
|
|
os.chown(destination, srcstat.st_uid, srcstat.st_gid)
|
|
except OSError: # lacking CAP_CHOWN, leave it owned by daemon
|
|
pass
|
|
filelist = []
|
|
for root, dirs, files in os.walk(targdir):
|
|
if not dirs and not files and root != targdir:
|
|
# otherwise empty directories have to be named to be synced
|
|
filelist.append(os.path.relpath(root, targdir))
|
|
for filename in files:
|
|
filelist.append(
|
|
os.path.relpath(os.path.join(root, filename), targdir))
|
|
await sshutil.prep_ssh_key('/etc/confluent/ssh/automation')
|
|
targip = node
|
|
if peerip:
|
|
targip = peerip
|
|
# Naming every file (--files-from implies --relative) together with
|
|
# --no-implied-dirs confines the preservation flags to the content
|
|
# actually being synchronized. Without it, the parent directories
|
|
# traversed on the way get the attributes of the staging copy, which
|
|
# is what forced preservation to be rolled back in the past.
|
|
output, stderr = await util.check_output(
|
|
'rsync', '-vLpgotDA', '--no-implied-dirs', '--files-from=-',
|
|
# content out of confluent's own storage is owned by the account
|
|
# the daemon runs as, which normally does not exist on the node
|
|
# and would arrive as a meaningless numeric id; every other owner
|
|
# is still matched by name as usual
|
|
'--usermap={}:root'.format(os.getuid()),
|
|
'--groupmap={}:root'.format(os.getgid()),
|
|
targdir + '/', 'root@[{}]:/'.format(targip),
|
|
input=''.join([x + '\n' for x in filelist]).encode('utf8'))
|
|
except subprocess.CalledProcessError as e:
|
|
unreadablefiles = []
|
|
for root, dirnames, filenames in os.walk(targdir):
|
|
for filename in filenames:
|
|
filename = os.path.join(root, filename)
|
|
try:
|
|
with open(filename, 'r') as _:
|
|
pass
|
|
except OSError:
|
|
unreadablefiles.append(filename.replace(targdir, ''))
|
|
if unreadablefiles:
|
|
raise Exception("Syncing failed due to unreadable files: " + ','.join(unreadablefiles))
|
|
elif hasattr(e, 'stderr') and e.stderr and b'Permission denied, please try again.' in e.stderr:
|
|
raise Exception('Syncing failed due to authentication error, is the confluent automation key not set up (osdeploy initialize -a) or is there some process replacing authorized_keys on the host?')
|
|
elif hasattr(e, 'stderr') and e.stderr:
|
|
sys.stderr.write(e.stderr.decode('utf8'))
|
|
raise
|
|
else:
|
|
raise
|
|
finally:
|
|
for root, dirnames, filenames in os.walk(targdir):
|
|
os.chown(root, os.getuid(), -1)
|
|
shutil.rmtree(targdir)
|
|
if not isinstance(output, str):
|
|
output = output.decode('utf8')
|
|
retval = {
|
|
'options': sl.optmap,
|
|
'output': output,
|
|
}
|
|
return retval # need dictionary with output and options
|
|
|
|
def stage_ent(currmap, ent, targdir, appendexist=False, correctionlist=None):
|
|
dst = currmap[ent]
|
|
everyfent = []
|
|
allfents = ent.split()
|
|
for tmpent in allfents:
|
|
fents = glob.glob(tmpent) # TODO: recursive globbing?
|
|
if not fents:
|
|
raise Exception('No matching files for "{}"'.format(tmpent))
|
|
everyfent.extend(fents)
|
|
if not everyfent:
|
|
raise Exception('No matching files for "{}"'.format(ent))
|
|
if dst is None: # this is to indicate source and destination as one
|
|
dst = os.path.dirname(everyfent[0]) + '/'
|
|
while dst and dst[0] == '/':
|
|
dst = dst[1:]
|
|
if len(everyfent) > 1 and dst[-1] != '/':
|
|
raise Exception(
|
|
'Multiple files match {}, {} needs a trailing slash to indicate a directory'.format(ent, dst))
|
|
fulltarg = os.path.join(targdir, dst)
|
|
for targ in everyfent:
|
|
mkpathorlink(targ, fulltarg, appendexist, correctionlist)
|
|
|
|
def mkpathorlink(source, destination, appendexist=False, correctionlist=None):
|
|
if os.path.isdir(source):
|
|
mkdirp(destination)
|
|
# Files are staged as symlinks, so rsync reads their attributes
|
|
# through to the real file, but a directory is staged as a directory
|
|
# and has to be given the source attributes to pass them along.
|
|
srcstat = os.stat(source)
|
|
if correctionlist is not None:
|
|
correctionlist.append((source, destination, srcstat))
|
|
for ent in os.listdir(source):
|
|
currsrc = os.path.join(source, ent)
|
|
currdst = os.path.join(destination, ent)
|
|
mkpathorlink(currsrc, currdst, appendexist, correctionlist)
|
|
else:
|
|
if destination[-1] == '/':
|
|
mkdirp(destination)
|
|
destination = os.path.join(destination, os.path.basename(source))
|
|
else:
|
|
mkdirp(os.path.dirname(destination))
|
|
if appendexist and os.path.exists(destination):
|
|
tmphdl, tmpnam = tempfile.mkstemp()
|
|
try:
|
|
shutil.copy(destination, tmpnam)
|
|
finally:
|
|
os.close(tmphdl)
|
|
os.remove(destination)
|
|
with open(destination, 'w') as realdest:
|
|
with open(tmpnam) as olddest:
|
|
realdest.write(olddest.read())
|
|
with open(source) as sourcedata:
|
|
realdest.write(sourcedata.read())
|
|
os.remove(tmpnam)
|
|
else:
|
|
if os.path.islink(destination):
|
|
os.remove(destination)
|
|
os.symlink(source, destination)
|
|
|
|
|
|
syncrunners = {}
|
|
cleaner = None
|
|
|
|
def start_syncfiles(nodename, cfg, suffixes, principals=[]):
|
|
global cleaner
|
|
peerip = None
|
|
if 'myips' in suffixes:
|
|
targips = suffixes['myips']
|
|
del suffixes['myips']
|
|
for targip in targips:
|
|
if targip in principals:
|
|
peerip = targip
|
|
break
|
|
deployinfo = cfg.get_node_attributes(
|
|
nodename, ('deployment.*',))
|
|
deployinfo = deployinfo.get(nodename, {})
|
|
profile = deployinfo.get(
|
|
'deployment.pendingprofile', {}).get('value', '')
|
|
if not profile:
|
|
profile = deployinfo.get(
|
|
'deployment.stagedprofile', {}).get('value', '')
|
|
if not profile:
|
|
profile = deployinfo.get(
|
|
'deployment.profile', {}).get('value', '')
|
|
if not profile:
|
|
raise Exception('Cannot perform syncfiles without profile assigned')
|
|
synclist = '/var/lib/confluent/public/os/{}/syncfiles'.format(profile)
|
|
if not os.path.exists(synclist):
|
|
return 200, 'OK', 'No synclist' # not running
|
|
sl = SyncList(synclist, nodename, cfg)
|
|
if not (sl.appendmap or sl.mergemap or sl.replacemap or sl.appendoncemap):
|
|
return 200, 'OK', 'Empty synclist' # the synclist has no actual entries
|
|
if nodename in syncrunners:
|
|
if syncrunners[nodename].done():
|
|
try:
|
|
syncrunners[nodename].result()
|
|
except Exception as e:
|
|
print(repr(e))
|
|
del syncrunners[nodename]
|
|
else:
|
|
return 503, 'Synchronization already in progress', 'Synchronization already in progress for {}'.format(nodename)
|
|
syncrunners[nodename] = tasks.spawn(
|
|
sync_list_to_node(sl, nodename, suffixes, peerip))
|
|
if not cleaner:
|
|
cleaner = tasks.spawn(cleanit())
|
|
return 202,'Queued', 'Background synchronization initiated' # backgrounded
|
|
|
|
|
|
async def cleanit():
|
|
toreap = {}
|
|
while True:
|
|
for nn in list(syncrunners):
|
|
if syncrunners[nn].done():
|
|
if nn in toreap:
|
|
try:
|
|
syncrunners[nn].result()
|
|
except Exception as e:
|
|
print(repr(e))
|
|
pass
|
|
del syncrunners[nn]
|
|
del toreap[nn]
|
|
else:
|
|
toreap[nn] = 1
|
|
elif nn in toreap:
|
|
del toreap[nn]
|
|
await asyncio.sleep(30)
|
|
|
|
|
|
def get_syncresult(nodename):
|
|
if nodename not in syncrunners:
|
|
return 204, 'Not Running', ''
|
|
if not syncrunners[nodename].done():
|
|
return 200, 'OK', ''
|
|
try:
|
|
result = syncrunners[nodename].result()
|
|
except Exception as e:
|
|
# Report the failure rather than a success with no payload.
|
|
print(repr(e))
|
|
del syncrunners[nodename]
|
|
return 500, 'Error', {'error': str(e)}
|
|
del syncrunners[nodename]
|
|
return 200, 'OK', result
|