X-Git-Url: http://git.vrable.net/?a=blobdiff_plain;f=cleaner%2Fcleaner;h=58292e4ede17693a045151ce680310c22ba223c9;hb=bdb05ff1ad95ab25e13934f66c83452ef00119fe;hp=336537da3b50c7bacc5539c65a06893de6e02387;hpb=1d2357b567318ccc6172f4b9edbbadbeba713086;p=bluesky.git

diff --git a/cleaner/cleaner b/cleaner/cleaner
index 336537d..58292e4 100755
--- a/cleaner/cleaner
+++ b/cleaner/cleaner
@@ -13,7 +13,8 @@ import boto
 from boto.s3.key import Key
 
 # The BlueSky 'struct cloudlog_header' data type.
-HEADER_FORMAT = '<4sb16sQIII'
+HEADER_FORMAT = '<4s48sb16sQIII'
+HEADER_CRYPTBYTES = 48
 HEADER_MAGIC = 'AgI-'
 HEADER_SIZE = struct.calcsize(HEADER_FORMAT)
 
@@ -114,6 +115,9 @@ class S3Backend:
 class LogItem:
     """In-memory representation of a single item stored in a log file."""
 
+    def __init__(self):
+        self.cryptkeys = '\0' * HEADER_CRYPTBYTES
+
     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])
 
@@ -132,7 +136,8 @@ class LogItem:
         link_locs = ''.join(link_locs)
 
         header = struct.pack(HEADER_FORMAT,
-                             HEADER_MAGIC, ord(self.type), self.id, self.inum,
+                             HEADER_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
 
@@ -211,28 +216,29 @@ class UtilizationTracker:
 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:
         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]
+    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:
@@ -268,7 +274,7 @@ def parse_log(data, location=None):
     offset = 0
     while len(data) - offset >= HEADER_SIZE:
         header = struct.unpack_from(HEADER_FORMAT, data, offset)
-        size = HEADER_SIZE + sum(header[4:7])
+        size = HEADER_SIZE + sum(header[5:8])
         if header[0] != HEADER_MAGIC:
             print "Bad header magic!"
             break
@@ -395,7 +401,7 @@ def rewrite_inode(backend, inode_map, inum, log, copy_data=True):
     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()):
@@ -410,7 +416,10 @@ def run_cleaner(backend, inode_map, log):
     # 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)