I would think that the link count should be sufficient. I've attached a python module that we've used to do some brick level indexing and cleanup on our system. You could use it to generate a list of the .glusterfs links for the remaining good files on the disk. Then you could check your candidates for removal against the list before removing them.
On Aug 21, 2014 12:05 PM, "Branden Timm" <btimm@xxxxxxxxxxxxxxx> wrote:
We have a distributed volume, and had a rather large (> 50TB) folder that was no longer needed. Naively, we removed the folder from the brick instead of through the Gluster client.
You can probably see where this is going. We didn’t actually reclaim the space because of the hard links to .glusterfs, and now we need to figure out how to clean up.
Is it sufficient to simply check whether a file under .glusterfs has less than 2 hard links, something like:
find .glusterfs -type f -links -2 -exec rm {} \;
Or do we have to do something else? Any help is much appreciated.
—Branden
_______________________________________________
Gluster-users mailing list
Gluster-users@xxxxxxxxxxx
http://supercolony.gluster.org/mailman/listinfo/gluster-users
import os
import os.path
import subprocess
from hashlib import md5
from socket import gethostname
from datetime import datetime
class File:
def __init__(self,path,brick=None):
self.path = os.path.realpath(path)
self.stat = os.stat(self.path)
self.brick = brick
self.xattrs = dict()
self._md5 = None
self._contains_nulls = None
self._all_nulls = None
def read_xattr(self,attrname):
if attrname in self.xattrs:
return self.xattrs[attrname]
p = subprocess.Popen(['getfattr', '--only-values', '-m', attrname, '-d', self.path],stdout=subprocess.PIPE, stderr=subprocess.PIPE)
(stdout,stderr) = p.communicate()
if stdout == '':
value = None
else:
value = stdout
self.xattrs[attrname] = value
return value
def is_sparse(self):
return self.stat.st_size > self.stat.st_blocks * 512
def is_linkfile(self):
if self.read_xattr('trusted.glusterfs.dht.linkto') != None:
return True
else:
return False
@property
def is_null(self):
if self._all_nulls == None:
self.__fastallnullcheck()
return self._all_nulls
def __fastallnullcheck(self):
"""
Tests to see if a file only contains nulls, exits as soon as a non null character is found.
This can also be done with __readfile, though it will read the full file.
"""
self._contains_nulls = False
if self.size > 0:
nullfile = True
f = open(self.path,"r")
for chunk in iter(lambda: f.read(512*2**10), ''):
for byte in chunk:
if ord(byte) != 0:
nullfile = False
break
else:
self._contains_nulls = True
if not nullfile:
break
f.close()
else:
nullfile = False
self._all_nulls = nullfile
def __readfile(self):
h = md5()
self._contains_nulls = False
if self.size > 0:
nullfile = True
f = open(self.path,"r")
for chunk in iter(lambda: f.read(512*2**10), ''):
h.update(chunk)
for byte in chunk:
if ord(byte) != 0:
nullfile = False
else:
self._contains_nulls = True
f.close()
else:
nullfile = False
self._all_nulls = nullfile
self._md5 = h.hexdigest()
@property
def contains_nulls(self):
if self._contains_nulls == None:
self.__readfile()
return self._contains_nulls
@property
def gfid(self):
gfid_raw = self.read_xattr("trusted.gfid")
if gfid_raw == None:
return None
gfid_array = ""
for i in bytearray(gfid_raw):
gfid_array += "%02x" % i
return gfid_array
@property
def gfidlinkpath(self):
gfid = self.gfid
if gfid == None or self.brick == None:
return None
relpath = os.path.join(".glusterfs",gfid[0:2],gfid[2:4],"%s-%s-%s-%s-%s" % (gfid[0:8],gfid[8:12],gfid[12:16],gfid[16:20],gfid[20:32]))
return os.path.join(self.brick.path,relpath)
@property
def linkcount(self):
return self.stat.st_nlink
@property
def size(self):
return self.stat.st_size
@property
def spaceallocated(self):
return self.stat.st_blocks * 512
@property
def relpath(self):
if self.brick != None:
return self.path[len(self.brick.path)+1:]
else:
return self.path
@property
def mtime(self):
""" Return mtime, defaults to float, I'm okay with that"""
return self.stat.st_mtime
@property
def md5sum(self):
if self._md5 == None:
self.__readfile()
return self._md5
def __str__(self):
self.__readfile()
if self.is_sparse():
sparsechar = 'S'
else:
sparsechar = '-'
if self.is_linkfile():
linkchar = 'L'
else:
linkchar = '-'
if self.is_null:
nullchar = 'N'
elif self.contains_nulls:
nullchar = 'n'
else:
nullchar = '-'
mtime = datetime.fromtimestamp(self.mtime)
return "%s %s %12d %3d %s %s %s" % (self.gfid,self.md5sum,self.size,self.linkcount,sparsechar + linkchar + nullchar,mtime,self.path)
class Brick:
def __init__(self,brickpath):
self._brickpath = os.path.realpath(brickpath)
@property
def node(self):
nodename = gethostname().split(".")[0]
return nodename
@property
def device(self):
return self.path.split(os.sep)[2]
@property
def path(self):
return self._brickpath
def __str__(self):
return "%s\t%s" % (self.node,self.path)
def __iter__(self):
for root,dirs,files in os.walk(self.path):
for file in files:
if not os.path.islink(os.path.join(root,file)):
f = File(os.path.join(root,file),brick=self)
yield f
_______________________________________________ Gluster-users mailing list Gluster-users@xxxxxxxxxxx http://supercolony.gluster.org/mailman/listinfo/gluster-users
