# Copyright (C) 2010 The Regents of the University of California
# Written by Michael Vrable <mvrable@cs.ucsd.edu>
-import base64, os, re, struct, sys
+import base64, os, re, struct, sys, time
import boto
+from boto.s3.key import Key
# The BlueSky 'struct cloudlog_header' data type.
-HEADER_FORMAT = '<4sb16sQIII'
-HEADER_MAGIC = 'AgI-'
+HEADER_FORMAT = '<4s48sb16sQIII'
+HEADER_CRYPTBYTES = 48
+HEADER_MAGIC1 = 'AgI-' # Unencrypted data
+HEADER_MAGIC2 = 'AgI=' # Encrypted data
HEADER_SIZE = struct.calcsize(HEADER_FORMAT)
+CHECKPOINT_MAGIC = struct.pack('<Q', 0x7ad7dafb42a498b4)
+
class ITEM_TYPE:
DATA = '1'
INODE = '2'
INODE_MAP = '3'
CHECKPOINT = '4'
-class FileBackend:
+class Backend:
+ """Base class for BlueSky storage backends."""
+
+ def loc_to_name(self, location):
+ return "log-%08d-%08d" % (location)
+
+ def name_to_loc(self, name):
+ m = re.match(r"^log-(\d+)-(\d+)$", name)
+ if m: return (int(m.group(1)), int(m.group(2)))
+
+ def dump_stats(self):
+ pass
+
+class FileBackend(Backend):
"""An interface to BlueSky where the log segments are on local disk.
This is mainly intended for testing purposes, as the real cleaner would
def __init__(self, path):
self.path = path
- def list(self):
+ def list(self, directory=0):
"""Return a listing of all log segments and their sizes."""
- files = [f for f in os.listdir(self.path) if f.startswith('log-')]
+ prefix = "log-%08d-" % (directory,)
+ files = [f for f in os.listdir(self.path) if f.startswith(prefix)]
files.sort()
return [(f, os.stat(os.path.join(self.path, f)).st_size)
for f in files]
- def read(self, filename):
+ def read(self, filename, offset=0, length=None):
fp = open(os.path.join(self.path, filename), 'rb')
- return fp.read()
+ if offset > 0:
+ fp.seek(offset)
+ if length is None:
+ return fp.read()
+ else:
+ return fp.read(length)
def write(self, filename, data):
fp = open(os.path.join(self.path, filename), 'wb')
fp.write(data)
fp.close()
- def loc_to_name(self, location):
- return "log-%08d-%08d" % (location)
+ def delete(self, filename):
+ os.unlink(os.path.join(self.path, filename))
+
+def retry_wrap(method):
+ def wrapped(self, *args, **kwargs):
+ for retries in range(3):
+ try:
+ return method(self, *args, **kwargs)
+ except:
+ print >>sys.stderr, "S3 operation failed, retrying..."
+ self.connect()
+ time.sleep(1.0)
+ return method(self, *args, **kwargs)
+ return wrapped
+
+class S3Backend(Backend):
+ """An interface to BlueSky where the log segments are on in Amazon S3."""
+
+ def __init__(self, bucket, path='', cachedir="."):
+ self.bucket_name = bucket
+ self.path = path
+ self.cachedir = cachedir
+ self.cache = {}
+ for f in os.listdir(cachedir):
+ self.cache[f] = True
+ print "Initial cache contents:", list(self.cache.keys())
+ self.connect()
+ self.stats_get = [0, 0]
+ self.stats_put = [0, 0]
+
+ def connect(self):
+ self.conn = boto.connect_s3(is_secure=False)
+ self.bucket = self.conn.get_bucket(self.bucket_name)
+
+ def list(self, directory=0):
+ files = []
+ prefix = "log-%08d-" % (directory,)
+ for k in self.bucket.list(self.path + prefix):
+ files.append((k.key, k.size))
+ return files
+
+ @retry_wrap
+ def read(self, filename, offset=0, length=None):
+ if filename in self.cache:
+ fp = open(os.path.join(self.cachedir, filename), 'rb')
+ if offset > 0:
+ fp.seek(offset)
+ if length is None:
+ return fp.read()
+ else:
+ return fp.read(length)
+ else:
+ k = Key(self.bucket)
+ k.key = self.path + filename
+ data = k.get_contents_as_string()
+ fp = open(os.path.join(self.cachedir, filename), 'wb')
+ fp.write(data)
+ fp.close()
+ self.cache[filename] = True
+ self.stats_get[0] += 1
+ self.stats_get[1] += len(data)
+ if offset > 0:
+ data = data[offset:]
+ if length is not None:
+ data = data[0:length]
+ return data
+
+ @retry_wrap
+ def write(self, filename, data):
+ k = Key(self.bucket)
+ k.key = self.path + filename
+ k.set_contents_from_string(data)
+ self.stats_put[0] += 1
+ self.stats_put[1] += len(data)
+ if filename in self.cache:
+ del self.cache[filename]
+
+ @retry_wrap
+ def delete(self, filename):
+ k = Key(self.bucket)
+ k.key = self.path + filename
+ k.delete()
+ if filename in self.cache:
+ del self.cache[filename]
+
+ def dump_stats(self):
+ print "S3 statistics:"
+ print "GET: %d ops / %d bytes" % tuple(self.stats_get)
+ print "PUT: %d ops / %d bytes" % tuple(self.stats_put)
+
+class SimpleBackend(Backend):
+ """An interface to the simple BlueSky test network server."""
+
+ def __init__(self, server=('localhost', 12345), cachedir="."):
+ self.bucket_name = bucket
+ self.server_address = server
+ self.cachedir = cachedir
+ self.cache = {}
+
+ def _get_socket(self):
+ return socket.create_connection(self.server_address).makefile()
+
+ def list(self, directory=0):
+ files = []
+ prefix = "log-%08d-" % (directory,)
+ for k in self.bucket.list(self.path + prefix):
+ files.append((k.key, k.size))
+ return files
+
+ def read(self, filename, offset=0, length=None):
+ if filename in self.cache:
+ fp = open(os.path.join(self.cachedir, filename), 'rb')
+ if offset > 0:
+ fp.seek(offset)
+ if length is None:
+ return fp.read()
+ else:
+ return fp.read(length)
+ else:
+ f = self._get_socket()
+ f.write("GET %s %d %d\n" % (filename, 0, 0))
+ f.flush()
+ datalen = int(f.readline())
+ if datalen < 0:
+ raise RuntimeError
+ data = f.read(datalen)
+ fp = open(os.path.join(self.cachedir, filename), 'wb')
+ fp.write(data)
+ fp.close()
+ self.cache[filename] = True
+ if offset > 0:
+ data = data[offset:]
+ if length is not None:
+ data = data[0:length]
+ return data
- def name_to_loc(self, name):
- m = re.match(r"^log-(\d+)-(\d+)$", name)
- if m: return (int(m.group(1)), int(m.group(2)))
+ def write(self, filename, data):
+ f = self._get_socket()
+ f.write("PUT %s %d %d\n" % (filename, len(data)))
+ f.write(data)
+ f.flush()
+ result = int(f.readline())
+ if filename in self.cache:
+ del self.cache[filename]
+
+ def delete(self, filename):
+ pass
class LogItem:
"""In-memory representation of a single item stored in a log file."""
+ def __init__(self):
+ self.cryptkeys = '\0' * HEADER_CRYPTBYTES
+ self.encrypted = False
+
def __str__(self):
- return "<Item ty=%s location=%s size=%d id=%s...>" % (self.type, self.location, self.size, base64.b16encode(self.id).lower()[0:8])
+ return "<Item%s ty=%s location=%s size=%d id=%s...>" % (self.encrypted and '$' or '', self.type, self.location, self.size, base64.b16encode(self.id).lower()[0:8])
@staticmethod
def random_id():
link_ids = ''.join(link_ids)
link_locs = ''.join(link_locs)
+ if self.encrypted:
+ magic = HEADER_MAGIC2
+ else:
+ magic = HEADER_MAGIC1
header = struct.pack(HEADER_FORMAT,
- HEADER_MAGIC, ord(self.type), self.id, self.inum,
+ magic, self.cryptkeys,
+ ord(self.type), self.id, self.inum,
len(self.data), len(link_ids), len(link_locs))
return header + self.data + link_ids + link_locs
self.backend = backend
self.dir_num = dir
self.seq_num = 0
- for logname in backend.list():
+ for logname in backend.list(dir):
+ print "Old log file:", logname
loc = backend.name_to_loc(logname[0])
if loc is not None and loc[0] == dir:
self.seq_num = max(self.seq_num, loc[1] + 1)
def __init__(self, backend):
self.segments = {}
- for (segment, size) in backend.list():
+ for (segment, size) in backend.list(0) + backend.list(1):
self.segments[segment] = [size, 0]
def add_item(self, item):
def parse_item(data):
if len(data) < HEADER_SIZE: return
header = struct.unpack_from(HEADER_FORMAT, data, 0)
- size = HEADER_SIZE + sum(header[4:7])
+ size = HEADER_SIZE + sum(header[5:8])
- if header[0] != HEADER_MAGIC:
+ if header[0] not in (HEADER_MAGIC1, HEADER_MAGIC2):
print "Bad header magic!"
return
if len(data) != size:
- print "Item size does not match!"
+ print "Item size does not match: %d != %d" % (size, len(data))
return
item = LogItem()
- item.id = header[2]
- item.inum = header[3]
+ if header[0] == HEADER_MAGIC2: item.encrypted = True
+ item.cryptkeys = header[1]
+ item.id = header[3]
+ item.inum = header[4]
item.location = None
- item.type = chr(header[1])
+ item.type = chr(header[2])
item.size = size
- item.data = data[HEADER_SIZE : HEADER_SIZE + header[4]]
+ item.data = data[HEADER_SIZE : HEADER_SIZE + header[5]]
links = []
- link_ids = data[HEADER_SIZE + header[4]
- : HEADER_SIZE + header[4] + header[5]]
- link_locs = data[HEADER_SIZE + header[4] + header[5]
- : HEADER_SIZE + sum(header[4:7])]
+ link_ids = data[HEADER_SIZE + header[5]
+ : HEADER_SIZE + header[5] + header[6]]
+ link_locs = data[HEADER_SIZE + header[5] + header[6]
+ : HEADER_SIZE + sum(header[5:8])]
for i in range(len(link_ids) // 16):
id = link_ids[16*i : 16*i + 16]
if id == '\0' * 16:
The elements of the tuple are (directory, sequence, offset, size)."""
filename = backend.loc_to_name((location[0], location[1]))
- data = backend.read(filename)[location[2] : location[2] + location[3]]
+ data = backend.read(filename, location[2], location[3])
item = parse_item(data)
item.location = location
return item
offset = 0
while len(data) - offset >= HEADER_SIZE:
header = struct.unpack_from(HEADER_FORMAT, data, offset)
- size = HEADER_SIZE + sum(header[4:7])
- if header[0] != HEADER_MAGIC:
+ size = HEADER_SIZE + sum(header[5:8])
+ if header[0] not in (HEADER_MAGIC1, HEADER_MAGIC2):
print "Bad header magic!"
break
if size + offset > len(data):
if item is not None: yield item
offset += size
-def load_checkpoint_record(backend):
- for (log, size) in reversed(backend.list()):
+def load_checkpoint_record(backend, directory=0):
+ for (log, size) in reversed(backend.list(directory)):
for item in reversed(list(parse_log(backend.read(log), log))):
+ print item
if item.type == ITEM_TYPE.CHECKPOINT:
return item
This will also build up information about segment utilization."""
+ self.version_vector = {}
self.checkpoint_record = checkpoint_record
util = UtilizationTracker(backend)
inodes = {}
self.obsolete_segments = set()
+ data = checkpoint_record.data
+ if not data.startswith(CHECKPOINT_MAGIC):
+ raise ValueError, "Invalid checkpoint record!"
+ data = data[len(CHECKPOINT_MAGIC):]
+ (vvlen,) = struct.unpack_from("<I", data, 0)
+ self.vvsize = 4 + 8*vvlen
+ for i in range(vvlen):
+ (v1, v2) = struct.unpack_from("<II", data, 4 + 8*i)
+ self.version_vector[v1] = v2
+ print self.version_vector
+ self.version_vector[checkpoint_record.location[0]] \
+ = checkpoint_record.location[1]
+ print self.version_vector
+
+ data = data[self.vvsize:]
+
print "Inode map:"
- for i in range(len(checkpoint_record.data) // 16):
- (start, end) = struct.unpack_from("<QQ", checkpoint_record.data, 16*i)
+ for i in range(len(data) // 16):
+ (start, end) = struct.unpack_from("<QQ", data, 16*i)
imap = load_item(backend, checkpoint_record.links[i][1])
util.add_item(imap)
print "[%d, %d]: %s" % (start, end, imap)
print
print "Segment utilizations:"
+ total_data = [0, 0]
+ deletions = [0, 0]
for (s, u) in sorted(util.segments.items()):
- #if u[1] > 0:
+ for i in range(2): total_data[i] += u[i]
print "%s: %s %s" % (s, u, float(u[1]) / u[0])
+ if u[1] == 0:
+ print "Would delete..."
+ backend.delete(s)
+ deletions[0] += 1
+ deletions[1] += u[0]
self.inodes = inodes
self.util = util
self.updated_inodes = set()
+ print "%d bytes total / %d bytes used" % tuple(total_data)
+ print "would delete %d segments (%d bytes)" % tuple(deletions)
+
def mark_updated(self, inum):
self.updated_inodes.add(inum)
new_checkpoint.id = LogItem.random_id()
new_checkpoint.inum = 0
new_checkpoint.type = ITEM_TYPE.CHECKPOINT
- new_checkpoint.data = ""
+ new_checkpoint.data = CHECKPOINT_MAGIC
new_checkpoint.links = []
- for i in range(len(self.checkpoint_record.data) // 16):
- (start, end) = struct.unpack_from("<QQ", self.checkpoint_record.data, 16*i)
+ new_checkpoint.data += struct.pack('<I', len(self.version_vector))
+ for d in sorted(self.version_vector):
+ new_checkpoint.data += struct.pack('<II', d, self.version_vector[d])
+
+ data = self.checkpoint_record.data[self.vvsize + len(CHECKPOINT_MAGIC):]
+ for i in range(len(data) // 16):
+ (start, end) = struct.unpack_from("<QQ", data, 16*i)
- new_checkpoint.data += self.checkpoint_record.data[16*i : 16*i + 16]
+ new_checkpoint.data += data[16*i : 16*i + 16]
# Case 1: No inodes in this range of the old inode map have
# changed. Simply emit a new pointer to the same inode map block.
log.write(new_checkpoint, 2)
self.checkpoint_record = new_checkpoint
-def rewrite_inode(backend, inode_map, inum, log):
+def rewrite_inode(backend, inode_map, inum, log, copy_data=True):
inode = inode_map.inodes[inum]
- blocks = []
- for l in inode.links:
- data = load_item(backend, l[1])
- blocks.append(data)
- log.write(data, 0)
- inode.links = [(b.id, b.location) for b in blocks]
+ if copy_data:
+ blocks = []
+ for l in inode.links:
+ data = load_item(backend, l[1])
+ blocks.append(data)
+ log.write(data, 0)
+ inode.links = [(b.id, b.location) for b in blocks]
log.write(inode, 1)
inode_map.mark_updated(inum)
-def run_cleaner(backend, inode_map, log):
+def run_cleaner(backend, inode_map, log, repack_inodes=False):
# Determine which segments are poorly utilized and should be cleaned. We
# need better heuristics here.
for (s, u) in sorted(inode_map.util.segments.items()):
- if float(u[1]) / u[0] < 0.99 and u[1] > 0:
+ if (float(u[1]) / u[0] < 0.6) and u[1] > 0:
print "Should clean segment", s
loc = backend.name_to_loc(s)
if s: inode_map.obsolete_segments.add(loc)
+ # TODO: We probably also want heuristics that will find inodes with
+ # badly-fragmented data and rewrite that to achieve better locality.
+
# Given that list of segments to clean, scan through those segments to find
# data which is still live and mark relevant inodes as needing to be
# rewritten.
- dirty_inodes = set()
+ if repack_inodes:
+ dirty_inodes = set(inode_map.inodes)
+ else:
+ dirty_inodes = set()
+ dirty_inode_data = set()
for s in inode_map.obsolete_segments:
filename = backend.loc_to_name(s)
print "Scanning", filename, "for live data"
for item in parse_log(backend.read(filename), filename):
if item.type in (ITEM_TYPE.DATA, ITEM_TYPE.INODE):
if item.inum != 0:
- dirty_inodes.add(item.inum)
+ inode = inode_map.inodes[item.inum]
+ if s == inode.location[0:2]:
+ dirty_inodes.add(item.inum)
+ if item.inum not in dirty_inode_data:
+ for b in inode.links:
+ if s == b[1][0:2]:
+ dirty_inode_data.add(item.inum)
+ break
print "Inodes to rewrite:", dirty_inodes
- for i in sorted(dirty_inodes):
- rewrite_inode(backend, inode_map, i, log)
+ print "Inodes with data to rewrite:", dirty_inode_data
+ for i in sorted(dirty_inodes.union(dirty_inode_data)):
+ rewrite_inode(backend, inode_map, i, log, i in dirty_inode_data)
if __name__ == '__main__':
- backend = FileBackend(".")
+ start_time = time.time()
+ backend = S3Backend("mvrable-bluesky-west", cachedir="/tmp/bluesky-cache")
+ #backend = FileBackend(".")
chkpt = load_checkpoint_record(backend)
+ print backend.list()
imap = InodeMap()
imap.build(backend, chkpt)
print chkpt
- log_dir = LogDirectory(backend, 0)
+ log_dir = LogDirectory(backend, 1)
run_cleaner(backend, imap, log_dir)
+ print "Version vector:", imap.version_vector
imap.write(backend, log_dir)
log_dir.close_all()
+ end_time = time.time()
+ print "Cleaner running time:", end_time - start_time
+ backend.dump_stats()
projects / bluesky.git / blobdiff