[Zodb-checkins] SVN: ZODB/trunk/ Merge r26771 from 3.3 branch.

Tim Peters tim.one at comcast.net
Mon Jul 26 21:35:50 EDT 2004

Log message for revision 26772:
  Merge r26771 from 3.3 branch.
  repozo improvement suggested by Toby Dickenson.
  Write backups to a temp file first; flush and fsync when done;
  only then rename; in case of a crash, no relevant damaged file
  will be left behind then.  And on Windows, Python's os.fsync()
  is necessary to convince the OS to write anything to disk.

  U   ZODB/trunk/NEWS.txt
  U   ZODB/trunk/src/scripts/repozo.py

Modified: ZODB/trunk/NEWS.txt
--- ZODB/trunk/NEWS.txt	2004-07-27 01:30:35 UTC (rev 26771)
+++ ZODB/trunk/NEWS.txt	2004-07-27 01:35:50 UTC (rev 26772)
@@ -35,6 +35,14 @@
+repozo.py:  Thanks to a suggestion from Toby Dickenson, backups
+(whether incremental or full) are first written to a temp file now,
+which is fsync'ed at the end, and only after that succeeds is the
+file renamed to YYYY-MM-DD-HH-MM-SS.ext form.  In case of a system
+crash during a repozo backup, this at least makes it much less
+likely that a backup file with incomplete or incorrect data will be
+left behind.
 fsrefs.py:  Fleshed out the module docstring, and repaired a bug
 wherein spurious error msgs could be produced after reporting a
 problem with an unloadable object.

Modified: ZODB/trunk/src/scripts/repozo.py
--- ZODB/trunk/src/scripts/repozo.py	2004-07-27 01:30:35 UTC (rev 26771)
+++ ZODB/trunk/src/scripts/repozo.py	2004-07-27 01:35:50 UTC (rev 26772)
@@ -177,6 +177,14 @@
     return options
+# afile is a Python file object, or created by gzip.open().  The latter
+# doesn't have a fileno() method, so to fsync it we need to reach into
+# its underlying file object.
+def fsync(afile):
+    afile.flush()
+    fileobject = getattr(afile, 'fileobj', afile)
+    os.fsync(fileobject.fileno())
 # Read bytes (no more than n, or to EOF if n is None) in chunks from the
 # current position in file fp.  Pass each chunk as an argument to func().
 # Return the total number of bytes read == the total number of bytes
@@ -211,21 +219,28 @@
 def copyfile(options, dst, start, n):
     # Copy bytes from file src, to file dst, starting at offset start, for n
-    # length of bytes
+    # length of bytes.  For robustness, we first write, flush and fsync
+    # to a temp file, then rename the temp file at the end.
     sum = md5.new()
     ifp = open(options.file, 'rb')
+    tempname = os.path.join(os.path.dirname(dst), 'tmp.tmp')
     if options.gzip:
-        ofp = gzip.open(dst, 'wb')
+        ofp = gzip.open(tempname, 'wb')
-        ofp = open(dst, 'wb')
+        ofp = open(tempname, 'wb')
     def func(data):
     ndone = dofile(func, ifp, n)
+    assert ndone == n
+    ifp.close()
+    fsync(ofp)
-    ifp.close()
-    assert ndone == n
+    os.rename(tempname, dst)
     return sum.hexdigest()
@@ -353,6 +368,8 @@
     datfile = os.path.splitext(dest)[0] + '.dat'
     fp = open(datfile, 'w')
     print >> fp, dest, 0, pos, sum
+    fp.flush()
+    os.fsync(fp.fileno())
@@ -381,6 +398,8 @@
     # This .dat file better exist.  Let the exception percolate if not.
     fp = open(datfile, 'a')
     print >> fp, dest, reposz, pos, sum
+    fp.flush()
+    os.fsync(fp.fileno())

More information about the Zodb-checkins mailing list