Hot File Clustering deactivated in 10.6
apple Verified ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
It seems that Hot File Clustering is deactivated since the Snow Leopard Upgrade.
I´ve deleted the .hotfiles.btree File, but the "clustering stage" is still "HFC_DISABLED".
# Hot File Clustering
clustering stage = HFC_DISABLED
recording period start time = Thu Jan 1 00:00:00 1970
recording period stop time = Thu Jan 1 00:00:00 1970
opaque recording data = 0x0
maximum files to track = 0
vnode of Hot Files B-Tree = 0x0
# Metadata Zone
metadata zone start block = 0
metadata zone end block = 0
hotfile start block = 0
hotfile end block = 0
hotfile free blocks = 0
hotfile maximum blocks = 0
overflow maximum blocks = 0
catalog maximum blocks = 0
Any idea?
Thanks in advance
15 replies
Sign in or join Beachball Help to reply.
- 0
Charles E. Flynn ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
If you explain to people how you got the data you posted, someone else can check to see if they get similar results.
- 0
R C-R ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
According to
this manual, the "HFC_DISABLED" clustering stage occurs most often because the volume is not the root volume (IOW, the startup volume) or it was unmounted while HFC was in its recording stage (the HFC_RECORDINGHFC stage).Also note that Hot File Clustering is disabled for volumes mounted as read only, for volumes less than 10 GB in size, & if HFS+ Journaling is disabled.
- 0
schnaufifant ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
I´ve used hfsdebug ---> http://osxbook.com/software/hfsdebug/
- 0
schnaufifant ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
Thanks for Responding (and the Link).
The Volume is the Startup Volume and the Volume Size is 50 GB.
With Tiger and Leopard (on the same Partition) the Hot File Clustering worked. - 0
KJK555 ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
##
Betsy7:~ kj$ sudo hfsdebug -b hotfile
# HFS+ Hot File Clustering (HFC) B-Tree
# B-Tree Node Descriptor
fLink = 0
bLink = 0
kind = 1 (kBTHeaderNode)
height = 0
numRecords = 3
reserved = 0
# B-Tree Header Record
treeDepth = 2
rootNode = 3
leafRecords = 5124
firstLeafNode = 13
lastLeafNode = 1
nodeSize = 4096 bytes
maxKeyLength = 10 bytes
totalNodes = 48
freeNodes = 12
reserved1 = 0
clumpSize = 0 (ignored)
btreeType = 128 (kUserBTreeType)
keyCompareType = 0 (unspecified/default)
attributes = 00000000000000000000000000000010
. kBTBigKeys (keyLength is UInt16)# User Data Record
magic = 0XFF28FF26
version = 1
duration = 216000 seconds
timebase = Fri Nov 6 00:06:49 2009
timeleft = 39837 seconds
threshold = 24
maxfileblks = 2560 blocks
maxfilecnt = 1000
tag = CLUSTERED HOT FILES B-TREE
##Mine appears to be intact, if I'm reading it right.
Kj ♘ - 0
R C-R ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
The Volume is the Startup Volume and the Volume Size is 50 GB.
Have you verified that this volume is journaled? What is the partition scheme for its HD?
- 0
KJK555 ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
After a bit of research, I suspect a bad block or a corrupt .hotfiles.btree file. Hot file clustering
does depend on journaling being on, so use Disk Utility to temporarily turn off journaling. If the
problem arose with adaptive hotfile clustering, deleting the invisible file at the root level of the
drive, .hotfiles.btree, will let a new .hotfiles.btree file be created when you turn journaling back on.Run a Surface Scan to check for unremapped bad blocks after the problem is solved.
This may require a third party disk utility such as tech tool.
If you do not have such a tool, backup your boot drive (clone or time machine) installation,
erase your drive using at minimum the "Zero Out Data" option. This will force the disk controller
on the boot drive to remap any bad sectors that it may encounter while erasing data.Afterwards, restore your OS Backup.
If you continue to have any other bad sector related problems, then it will be time for a new
Hard Drive.
As I always say: "A Hard Drive is a failure waiting to Happen".
Kj ♘ - 0
R C-R ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
KJK555 wrote:
If you do not have such a tool, backup your boot drive (clone or time machine) installation,
erase your drive using at minimum the "Zero Out Data" option. This will force the disk controller
on the boot drive to remap any bad sectors that it may encounter while erasing data.In general, the "Zero Out Data" option does
not force the drive to remap any bad sectors while writing zeros to the drive. This only happens when the drive
reads data & detects an unrecoverable checksum error or read after write (RAW) is turned on. On all modern consumer level ATA drives, RAW is automatically turned off after the first few hours of use because it nearly halves drive performance.The "Zero Out Data" option does not force the drive to read any sectors (again, because this would nearly double the time it takes to complete this already lengthy operation), so unless RAW is turned on, no bad sectors will be detected.
- 0
schnaufifant ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
The Partition Scheme is GUID, and Journaling is activated.
- 0
schnaufifant ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
Really strange: HFC seems to be deaktivated, but my timeleft Counter is counting backward.
sudo /Applications/Utilities/hfsdebug -b hotfile
# HFS+ Hot File Clustering (HFC) B-Tree
# B-Tree Node Descriptor
fLink = 0
bLink = 0
kind = 1 (kBTHeaderNode)
height = 0
numRecords = 3
reserved = 0
# B-Tree Header Record
treeDepth = 0
rootNode = 0
leafRecords = 0
firstLeafNode = 0
lastLeafNode = 0
nodeSize = 4096 bytes
maxKeyLength = 10 bytes
totalNodes = 16
freeNodes = 15
reserved1 = 0
clumpSize = 65536 (ignored)
btreeType = 128 (kUserBTreeType)
keyCompareType = 0 (unspecified/default)
attributes = 00000000000000000000000000000010
. kBTBigKeys (keyLength is UInt16)
# User Data Record
magic = 0XFF28FF26
version = 1
duration = 216000 seconds
timebase = Tue Nov 17 13:01:43 2009
timeleft = 202650 seconds
threshold = 24
maxfileblks = 2560 blocks
maxfilecnt = 1000
tag = CLUSTERED HOT FILES B-TREE - 0
KJK555 ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
"The process of erasing a disk, partition, or volume by writing zeros to every bit on such is called
zeroing. Zeroing finds bad sectors and +maps them out of service+, also known as
sparing.
When an attempt to write zeros to bad sectors fails, the bad sectors are both marked as occupied in
the directory and added to the +bad blocks file+ of the file system. Once the bad sectors have been
spared, no attempt will ever be made to read-from or write-to them again. For details, see Apple
Developer Connection Technical Note TN1150, "HFS Plus Volume Format."Taken from:
http://www.thexlab.com/faqs/repairprocess.html#Anchor-Reinstall-49575
Kj ♘ - 0
R C-R ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
With all due respect for the XLab's FAQ's, it is physically impossible for a conventional hard drive to detect a bad sector during a
write operation. A bad sector can only be detected during a
read operation, when the drive either detects an anomaly in the raw analog signal coming off the read head or a checksum error in the digital data that signal is converted to. This is not possible during a write because the head is busy applying the strong magnetic field to the platter that physically writes the data to it. IOW, the head can read or write, but not do both at the same time.As previously mentioned, only if RAW is enabled will the drive
itself automatically force a read operation immediately following a write one (while the data to be written is still in the drive's buffer & thus can be rewritten to a good sector if it is necessary to spare out the original one). Typically, this read occurs on the next revolution of the platter, before the head moves to a different track, so there is no additional performance penalty from an additional seek. However, it should be obvious that there is still a performance penalty because each operation takes a minimum of two revolutions of the platter instead of just one.Because of this, regular consumer level ATA hard drives do not leave RAW enabled indefinitely. It is typical for these drives to be shipped with RAW enabled as a safeguard against sector failures that are most likely to occur very early in the drives life (so called "infant mortality" failures due to initial manufacturing defects), but RAW is switched off automatically after a certain number of writes or hours of operation to avoid the performance penalty leaving it enabled permanently would otherwise incur. (It is for this reason that sites like Bare Feats "break in" hard drives before testing -- if they did not, new drives would test poorly.)
Some enterprise level drives do leave RAW enabled permanently, or switch it on & off for other reasons (see
this Seagate white paper pdf for a novel use of the technology), & many SCSI drives still allow the user to control RAW through configuration of mode pages, but few users are using these drives.Of course, as
Technical Note TN1150 mentions, it is still possible for the OS to spare out bad sectors on HFS+ volumes by use of the extent overflow file in the file system of the volume, but this is not something the drive does by itself, & it still requires a read to detect any bad sectors.So, in the most general case (a not new, consumer grade ATA drive), it boils down to this: Does the built-in "Zero data" option in the OS actually do a read of the sectors it writes zeros to, or does it just write the pattern without verifying it afterwards?
To be honest, I can't find a definitive statement about this; however, consider that for its intended purpose as a security erase this is clearly not necessary or desirable (since it nearly doubles the time the operation would take), & more tellingly it would be pointless to write spared out sector data to an HFS extent overflow file that is itself going to be erased. I've also done some testing with various ATA drives, comparing the time it actually takes to do a zero data erase with calculations of how long it should take based on the published read & write specs of the drives & the size of the volume, & there is no indication I can find from this that the OS actually does take the time to do a read verification.
Finally, note that AFAIK the above is consistent with how the "Surface Scan" utilities differ from the zero data option of Disk Utility: they can be designed to do read verifications or other tests of sector reliability & mark sectors as bad in the extent overflow file independently of any 'in drive' checks, but obviously this requires that the file system itself not be erased (& that file along with it), so it isn't quite the same thing.
- 0
KJK555 ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
Hi R C -R:
I'm still of the belief that one should have the necessary tools on hand to do hard drive repairs
if their data has any value at all to them. Disk Utility, as good as it is, is not a complete solution
by itself. One should own at least one other good third party utility such as Disk Warrior, Tech
Tool Pro or Drive Genius to find/fix many common problems that Disk Utility won't find/fix.Kj ♘
- 0
R C-R ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
Certainly, one needs the right tool for the job, but the point I was trying to make is that a bad block would not explain the deactivation of hot file clustering: the drive would detect a bad block in the .hotfiles.btree file as soon as it tried to read from it & return a read error to the OS.
A lot of users seem to have the impression that they need special utilities to reliably map out bad sectors, but this is not true for essentially all modern drives. As I mentioned, they continually monitor read operations for errors without the need of any help from the OS or additional utilities. When a "hard" (unrecoverable) error occur, data loss usually results, but the drive maps out the sector as bad & returns a read error -- it is not as if it just ignores the error, returns bogus data to the OS, & continues to use the bad sector.
It would be different if for some reason the OS is actually writing bad hot files data to the file: as far as the drive is concerned, the data would be good. I don't know if the OS would detect this or what it would do about it if it did, but I suppose it is conceivable that it would deactivate hot file clustering. If so, deleting the file as you suggested might fix things, but I suspect the ultimate remedy is to reinstall the OS to eliminate whatever is causing the OS to write bad data to the file to begin with.
- 0
V.K. ·
Imported from the Apple Support Forums via the Wayback Machine 🔗
R C-R wrote:
Certainly, one needs the right tool for the job, but the point I was trying to make is that a bad block would not explain the deactivation of hot file clustering: the drive would detect a bad block in the .hotfiles.btree file as soon as it tried to read from it & return a read error to the OS.A lot of users seem to have the impression that they need special utilities to reliably map out bad sectors, but this is not true for essentially all modern drives. As I mentioned, they continually monitor read operations for errors without the need of any help from the OS or additional utilities.
When a "hard" (unrecoverable) error occur, data loss usually results, but the drive maps out the sector as bad & returns a read error -- it is not as if it just ignores the error, returns bogus data to the OS, & continues to use the bad sector.sorry to butt in here - very interesting thread. could you clarify that point please? as I understand it, you are saying that bad sectors are mapped out on reads but not writes by the hard drive itself. why does this eliminate the need to map out bad sectors then? if some bad sectors exist that are not mapped out yet and the OS tries to write to them, the drive will not stop it, right? that will result in corrupt unrecoverable data, won't it? what am i missing here?
It would be different if for some reason the OS is actually writing bad hot files data to the file: as far as the drive is concerned, the data would be good. I don't know if the OS would detect this or what it would do about it if it did, but I suppose it is conceivable that it would deactivate hot file clustering. If so, deleting the file as you suggested might fix things, but I suspect the ultimate remedy is to reinstall the OS to eliminate whatever is causing the OS to write bad data to the file to begin with.