Add S3 backend for the cleaner.

It does not yet cache files so performance is poor.

diff --git a/cleaner/cleaner b/cleaner/cleaner
index b018d0d..4267b6e 100755 (executable)
--- a/cleaner/cleaner
+++ b/cleaner/cleaner
@@ -10,6 +10,7 @@
 
 import base64, os, re, struct, sys
 import boto
+from boto.s3.key import Key
 
 # The BlueSky 'struct cloudlog_header' data type.
 HEADER_FORMAT = '<4sb16sQIII'
@@ -59,6 +60,42 @@ class FileBackend:
         m = re.match(r"^log-(\d+)-(\d+)$", name)
         if m: return (int(m.group(1)), int(m.group(2)))
 
+class S3Backend:
+    """An interface to BlueSky where the log segments are on in Amazon S3."""
+
+    def __init__(self, bucket, path='', cachedir=None):
+        self.conn = boto.connect_s3(is_secure=False)
+        self.bucket = self.conn.get_bucket(bucket)
+        self.path = path
+
+    def list(self):
+        files = []
+        for k in self.bucket.list(self.path + 'log-'):
+            files.append((k.key, k.size))
+        return files
+
+    def read(self, filename):
+        k = Key(self.bucket)
+        k.key = self.path + filename
+        return k.get_contents_as_string()
+
+    def write(self, filename, data):
+        k = Key(self.bucket)
+        k.key = self.path + filename
+        k.set_contents_from_string(data)
+
+    def delete(self, filename):
+        k = Key(self.bucket)
+        k.key = self.path + filename
+        k.delete()
+
+    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)))
+
 class LogItem:
     """In-memory representation of a single item stored in a log file."""
 
@@ -232,6 +269,7 @@ def parse_log(data, location=None):
 def load_checkpoint_record(backend):
     for (log, size) in reversed(backend.list()):
         for item in reversed(list(parse_log(backend.read(log), log))):
+            print item
             if item.type == ITEM_TYPE.CHECKPOINT:
                 return item
 
@@ -346,7 +384,7 @@ def run_cleaner(backend, inode_map, log):
     # 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] < 0.6 and u[1] > 0:
+        if (float(u[1]) / u[0] < 0.6 or u[1] < 32768) and u[1] > 0:
             print "Should clean segment", s
             loc = backend.name_to_loc(s)
             if s: inode_map.obsolete_segments.add(loc)
@@ -380,8 +418,9 @@ def run_cleaner(backend, inode_map, log):
         rewrite_inode(backend, inode_map, i, log, i in dirty_inode_data)
 
 if __name__ == '__main__':
-    backend = FileBackend(".")
+    backend = S3Backend("mvrable-bluesky", cachedir=".")
     chkpt = load_checkpoint_record(backend)
+    print backend.list()
     imap = InodeMap()
     imap.build(backend, chkpt)
     print chkpt