1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
|
#!/usr/bin/env python
# -*- coding: utf-8 -*-
#
# Copyright (c) 2014-2015, NORDUnet A/S.
# See LICENSE for licensing information.
#
# Fetch new entries from all storage nodes.
# See catlfish/doc/merge.txt for more about the merge process.
#
import sys
import struct
import subprocess
import signal
import logging
from time import sleep
from multiprocessing import Process, Pipe
from random import Random
from mergetools import get_logorder, verify_entry, get_new_entries, \
chunks, fsync_logorder, get_entries, add_to_logorder, \
hexencode, hexdecode, parse_args, perm, flock_ex_or_fail, Status, \
terminate_child_procs
from certtools import timing_point, write_file, create_ssl_context
def merge_fetch_sequenced(args, config, localconfig):
paths = localconfig["paths"]
storagenodes = config["storagenodes"]
mergedb = paths["mergedb"]
logorderfile = mergedb + "/logorder"
statusfile = mergedb + "/merge_fetch.status"
s = Status(statusfile)
chainsdb = perm(localconfig.get("dbbackend", "filedb"), mergedb + "/chains")
own_key = (localconfig["nodename"],
"%s/%s-private.pem" % (paths["privatekeys"],
localconfig["nodename"]))
timing = timing_point()
logorder = get_logorder(logorderfile)
timing_point(timing, "get logorder")
certsinlog = set(logorder)
new_entries_per_node = {}
new_entries = set()
entries_to_fetch = {}
for storagenode in storagenodes:
logging.info("getting new entries from %s", storagenode["name"])
new_entries_per_node[storagenode["name"]] = \
set(get_new_entries(storagenode["name"],
"https://%s/" % storagenode["address"],
own_key, paths))
new_entries.update(new_entries_per_node[storagenode["name"]])
entries_to_fetch[storagenode["name"]] = []
timing_point(timing, "get new entries")
new_entries -= certsinlog
logging.info("adding %d entries", len(new_entries))
for ehash in new_entries:
for storagenode in storagenodes:
if ehash in new_entries_per_node[storagenode["name"]]:
entries_to_fetch[storagenode["name"]].append(ehash)
break
verifycert = subprocess.Popen(
[paths["verifycert_bin"], paths["known_roots"]],
stdin=subprocess.PIPE, stdout=subprocess.PIPE)
added_entries = 0
for storagenode in storagenodes:
nentries = len(entries_to_fetch[storagenode["name"]])
logging.info("getting %d entries from %s", nentries, storagenode["name"])
for chunk in chunks(entries_to_fetch[storagenode["name"]], 100):
entries = get_entries(storagenode["name"],
"https://%s/" % storagenode["address"],
own_key, paths, chunk)
for ehash in chunk:
entry = entries[ehash]
verify_entry(verifycert, entry, ehash)
chainsdb.add(ehash, entry)
add_to_logorder(logorderfile, ehash)
logorder.append(ehash)
certsinlog.add(ehash)
added_entries += 1
s.status("INFO: getting %d entries from %s: %d" %
(nentries, storagenode["name"], added_entries))
chainsdb.commit()
fsync_logorder(logorderfile)
timing_point(timing, "add entries")
logging.info("added %d entries", added_entries)
verifycert.communicate(struct.pack("I", 0))
if args.timing:
logging.debug("timing: merge_fetch: %s", timing["deltatimes"])
tree_size = len(logorder)
if tree_size == 0:
return (0, '')
else:
return (tree_size, logorder[tree_size-1])
def merge_fetch_worker(args, localconfig, storagenode, pipe):
paths = localconfig["paths"]
mergedb = paths["mergedb"]
chainsdb = perm(localconfig.get("dbbackend", "filedb"), mergedb + "/chains")
own_key = (localconfig["nodename"],
"%s/%s-private.pem" % (paths["privatekeys"],
localconfig["nodename"]))
to_fetch = set()
timeout = max(3, args.mergeinterval / 10)
while True:
if pipe.poll(timeout):
msg = pipe.recv().split()
if len(msg) < 2:
continue
cmd = msg[0]
ehash = msg[1]
if cmd == 'FETCH':
to_fetch.add(hexdecode(ehash))
else:
logging.warning("%s: unknown command from parent: %s",
storagenode["name"], msg)
if len(to_fetch) > 0:
logging.info("%s: fetching %d entries", storagenode["name"],
len(to_fetch))
# TODO: Consider running the verifycert process longer.
verifycert = subprocess.Popen(
[paths["verifycert_bin"], paths["known_roots"]],
stdin=subprocess.PIPE, stdout=subprocess.PIPE)
# Chunking for letting other workers take the chainsdb lock.
for chunk in chunks(list(to_fetch), 100):
chainsdb.lock_ex()
entries = get_entries(storagenode["name"],
"https://%s/" % storagenode["address"],
own_key, paths, chunk)
for ehash in chunk:
entry = entries[ehash]
verify_entry(verifycert, entry, ehash)
chainsdb.add(ehash, entry)
chainsdb.commit()
for ehash in chunk:
pipe.send('FETCHED %s' % hexencode(ehash))
to_fetch.remove(ehash)
verifycert.communicate(struct.pack("I", 0))
new_entries = get_new_entries(storagenode["name"],
"https://%s/" % storagenode["address"],
own_key, paths)
if len(new_entries) > 0:
logging.info("%s: got %d new entries", storagenode["name"],
len(new_entries))
for ehash in new_entries:
pipe.send('NEWENTRY %s' % hexencode(ehash))
def term(signal, arg):
terminate_child_procs()
sys.exit(1)
def newworker(name, args):
my_conn, child_conn = Pipe()
p = Process(target=merge_fetch_worker,
args=tuple(args + [child_conn]),
name='merge_fetch_%s' % name)
p.daemon = True
p.start()
logging.debug("%s started, pid %d", name, p.pid)
return (name, my_conn, p)
def merge_fetch_parallel(args, config, localconfig):
paths = localconfig["paths"]
storagenodes = config["storagenodes"]
mergedb = paths["mergedb"]
logorderfile = mergedb + "/logorder"
currentsizefile = mergedb + "/fetched"
rand = Random()
signal.signal(signal.SIGTERM, term)
procs = {}
for storagenode in storagenodes:
name = storagenode['name']
procs[name] = newworker(name, [args, localconfig, storagenode])
logorder = get_logorder(logorderfile) # List of entries in log.
entries_in_log = set(logorder) # Set of entries in log.
entries_to_fetch = set() # Set of entries to fetch.
fetch = {} # Dict with entries to fetch.
while procs:
assert(not entries_to_fetch)
# Poll worker processes.
for name, pipe, p in procs.values():
if not p.is_alive():
logging.warning("%s is gone, restarting", name)
procs[name] = newworker(name, [args, localconfig,
storagenodes[name]])
continue
logging.info("polling %s", name)
if pipe.poll(1):
msg = pipe.recv().split()
if len(msg) < 2:
logging.warning("unknown command from %s: %s", name, msg)
continue
cmd = msg[0]
ehash = msg[1]
if cmd == 'NEWENTRY':
logging.info("NEWENTRY at %s: %s", name, ehash)
entries_to_fetch.add(ehash)
logging.debug("entries_to_fetch: %s", entries_to_fetch)
elif cmd == 'FETCHED':
logging.info("FETCHED from %s: %s", name, ehash)
logorder.append(ehash)
add_to_logorder(logorderfile, hexdecode(ehash))
fsync_logorder(logorderfile)
entries_in_log.add(ehash)
if ehash in entries_to_fetch:
entries_to_fetch.remove(ehash)
del fetch[ehash]
else:
logging.warning("unknown command from %s: %s", name, msg)
# Ask workers to fetch entries.
logging.debug("nof entries to fetch including entries in log: %d",
len(entries_to_fetch))
entries_to_fetch -= entries_in_log
logging.info("entries to fetch: %d", len(entries_to_fetch))
# Add entries in entries_to_fetch as keys in dictionary fetch,
# values being a list of storage nodes, in randomised order.
for e in entries_to_fetch:
if not e in fetch:
l = procs.values()
rand.shuffle(l)
fetch[e] = l
# For each entry to fetch, treat its list of nodes as a
# circular list and ask the one in the front to fetch the
# entry.
while entries_to_fetch:
ehash = entries_to_fetch.pop()
nodes = fetch[ehash]
node = nodes.pop(0)
fetch[ehash] = nodes.append(node)
name, pipe, p = node
logging.info("asking %s to FETCH %s", name, ehash)
pipe.send("FETCH %s" % ehash)
# Update the 'fetched' file.
logsize = len(logorder)
if logsize == 0:
last_hash = ''
else:
last_hash = logorder[logsize - 1]
logging.info("updating 'fetched' file: %d %s", logsize-1, last_hash)
currentsize = {"index": logsize - 1, "hash": last_hash}
logging.debug("writing to %s: %s", currentsizefile, currentsize)
write_file(currentsizefile, currentsize)
return 0
def main():
"""
If no `--mergeinterval':
Fetch new entries from all storage nodes, in sequence, updating
the 'logorder' file and the 'chains' database.
Write 'fetched' to reflect how far in 'logorder' we've succesfully
fetched and verified.
If `--mergeinterval':
Start one process per storage node, read their stdout for learning
about two things: (i) new entries ready for fetching ("NEWENTRY") and
(ii) new entries being succesfully fetched ("FETCHED").
Write to their stdin ("FETCH") when they should fetch another entry.
Update 'logorder' and the 'chains' database as we see new FETCHED
messages.
Write 'fetched' to reflect how far in 'logorder' we've succesfully
fetched and verified.
Keep doing this forever.
NOTE: The point of having 'fetched' is that it can be atomically
written while 'logorder' cannot (unless we're fine with rewriting it
for each and every update, which we're not).
TODO: Deduplicate some code.
"""
args, config, localconfig = parse_args()
paths = localconfig["paths"]
mergedb = paths["mergedb"]
currentsizefile = mergedb + "/fetched"
lockfile = mergedb + "/.merge_fetch.lock"
loglevel = getattr(logging, args.loglevel.upper())
if args.mergeinterval is None:
logging.basicConfig(level=loglevel)
else:
logging.basicConfig(filename=args.logdir + "/merge_fetch.log",
level=loglevel)
if not flock_ex_or_fail(lockfile):
logging.critical("unable to take lock %s", lockfile)
return 1
create_ssl_context(cafile=paths["https_cacertfile"])
if args.mergeinterval:
return merge_fetch_parallel(args, config, localconfig)
else:
logsize, last_hash = merge_fetch_sequenced(args, config, localconfig)
currentsize = {"index": logsize - 1, "hash": hexencode(last_hash)}
logging.debug("writing to %s: %s", currentsizefile, currentsize)
write_file(currentsizefile, currentsize)
return 0
if __name__ == '__main__':
sys.exit(main())
|