Re: disk fragmentation, 0%?
Don't top-post, please. Lei Sun [EMAIL PROTECTED] writes: Then, my other question is, If the file space allocation works like Glenn said earlier, how come with the exact same files from 2 different installations using the exact procedures, can result in different fragmentation? in the atacontrol raid1 failure case, /dev/ar0s1a: ... 0.5% fragmentation in the new build case, /dev/ar0s1a: ... 0.0% fragmentation That doesn't seems to make a lot of sense. There are lots of possible explanations, including (non-harmful) race conditions. Is there some reason you care? ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation, 0%?
On Sun, Aug 14, 2005 at 01:30:41PM -0700, Glenn Dawson wrote: From the original message: Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp This shows that /tmp is empty. If the reserved space was being encroached upon, it would show 100% capacity, and available bytes would go negative, not bytes used. It would look something like this: Filesystem SizeUsed Avail Capacity Mounted on /dev/ad0s1a248M238M-10M 105%/ I've never seen the capacity go negative before, which is why I suggested someone else might know the answer. Ups, yes, that's really weird. It's so unusual that I didn't notice it the first time. Could that be some counter overflowing? -Glenn Regards, -cpghost. -- Cordula's Web. http://www.cordula.ws/ ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation, 0%?
Thanks for the good answers. But can anyone tell me why the capacity is going negative? and not full? Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp As someone mentioned, there is a FAQ on this. You should read it. It is going negative because you have used more than the nominal capacity of the slice. The nominal capacity is the total space minus the reserved proportion (usually 8%) that is held out. Root is able to write to that space and you have done something that got root to write beyond the nominal space. jerry Thanks a lot Lei On 8/14/05, Glenn Dawson [EMAIL PROTECTED] wrote: At 12:18 PM 8/14/2005, cpghost wrote: On Sun, Aug 14, 2005 at 12:09:19AM -0700, Glenn Dawson wrote: 2. How come /tmp is -0% in size? -278K? What had happened? as I have never experienced this in the previous installs on the exact same hardware. Not sure about that one. Maybe someone else has an answer. This is a FAQ. The available space is always computed after subtracting some space that would be only available to root (typically around 5% or 10% of the partition size). The default is 8%. This free space is necessary to avoid internal fragmentation and to keep the file system going. Root may be able to borrow some space from this (in which case the capacity goes below 0%), but it is not advisable to keep the file system so full, so it should be only for a limited period of time. The reason for having the reserved space is to allow the functions that allocate space to be able to find contiguous free space. When the disk is nearly full it takes longer and longer to locate contiguous space, which can lead to performance problems. In your example, you're 278K over the limit; and should delete some files to make space ASAP. Should /tmp fill up more, it will soon become inoperable. From the original message: Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp This shows that /tmp is empty. If the reserved space was being encroached upon, it would show 100% capacity, and available bytes would go negative, not bytes used. It would look something like this: Filesystem SizeUsed Avail Capacity Mounted on /dev/ad0s1a248M238M-10M 105%/ I've never seen the capacity go negative before, which is why I suggested someone else might know the answer. -Glenn ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation, 0%?
On 8/15/05, Jerry McAllister [EMAIL PROTECTED] wrote: As someone mentioned, there is a FAQ on this. You should read it. It is going negative because you have used more than the nominal capacity of the slice. The nominal capacity is the total space minus the reserved proportion (usually 8%) that is held out. Root is able to write to that space and you have done something that got root to write beyond the nominal space. I'm not sure you are right in this case. I think you need to re-read the post. I've quoted the relevent part here: Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp Looking at how the columns line up I have to state that I too have never seen this behaviour. As an experiment I over-filled a file system and here's the results: Filesystem SizeUsed Avail Capacity Mounted on /dev/ad0s1f965M895M -7.4M 101%/tmp Note capacity is not negative. So that makes three of us in this thread who have not seen negative capacity on UFS. I have seen negative capacity when running an old version of FreeBSD with a very large NFS mount (not enough bits in statfs if I remember correctly). jerry Frem. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation, 0%?
This happened, after I tested the atacontrol to rebuild the raid1. The /tmp partition doesn't have anything but several empty directories created. and I have the clear /tmp directive in the rc.conf, which will clean up the /tmp everytime when system boot up. So that was really wierd. as it never happened this way the previous time that I was rebuilding the raid1. Thanks Lei On 8/15/05, Freminlins [EMAIL PROTECTED] wrote: On 8/15/05, Jerry McAllister [EMAIL PROTECTED] wrote: As someone mentioned, there is a FAQ on this. You should read it. It is going negative because you have used more than the nominal capacity of the slice. The nominal capacity is the total space minus the reserved proportion (usually 8%) that is held out. Root is able to write to that space and you have done something that got root to write beyond the nominal space. I'm not sure you are right in this case. I think you need to re-read the post. I've quoted the relevent part here: Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp Looking at how the columns line up I have to state that I too have never seen this behaviour. As an experiment I over-filled a file system and here's the results: Filesystem SizeUsed Avail Capacity Mounted on /dev/ad0s1f965M895M -7.4M 101%/tmp Note capacity is not negative. So that makes three of us in this thread who have not seen negative capacity on UFS. I have seen negative capacity when running an old version of FreeBSD with a very large NFS mount (not enough bits in statfs if I remember correctly). jerry Frem. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation, 0%?
On Mon, Aug 15, 2005 at 09:20:12AM -0400, Jerry McAllister wrote: Thanks for the good answers. But can anyone tell me why the capacity is going negative? and not full? Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp As someone mentioned, there is a FAQ on this. You should read it. In fact, you're both wrong, because that's clearly not what's going on here (capacity 0, not capacity 100!) The only thing I can think of is that you have some filesystem corruption on this partition that is confusing the stats. Try dropping to single-user mode and running fsck -f /tmp. Kris pgpUKruIUuhAv.pgp Description: PGP signature
Re: disk fragmentation, 0%?
Thanks All, I think Kris's suggestion worked, as when I was rebuilding of the atacontrol, I remember it failed once, and had a lot of problem trying to reboot and unmount the /tmp directory. So after I rebuild the array, somehow /tmp looks clean to the OS, and didn't get checked. so somehow the the stats was not showing the correct information. I have already rebuild the machine, all of the effect from the atacontrol rebuild array are gone now, and it seems like everything is back to normal. Capacity is right, Used is right, Avail is right, and all 0.0% fragmentation. Then, my other question is, If the file space allocation works like Glenn said earlier, how come with the exact same files from 2 different installations using the exact procedures, can result in different fragmentation? in the atacontrol raid1 failure case, /dev/ar0s1a: ... 0.5% fragmentation in the new build case, /dev/ar0s1a: ... 0.0% fragmentation That doesn't seems to make a lot of sense. Thanks again Lei On 8/15/05, Kris Kennaway [EMAIL PROTECTED] wrote: On Mon, Aug 15, 2005 at 09:20:12AM -0400, Jerry McAllister wrote: Thanks for the good answers. But can anyone tell me why the capacity is going negative? and not full? Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp As someone mentioned, there is a FAQ on this. You should read it. In fact, you're both wrong, because that's clearly not what's going on here (capacity 0, not capacity 100!) The only thing I can think of is that you have some filesystem corruption on this partition that is confusing the stats. Try dropping to single-user mode and running fsck -f /tmp. Kris ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
disk fragmentation, 0%?
Hi, I know this question has been raised a lot of times, and most of people don't think it is necessary to defragment ufs, and from the previous posts, I got to know there are sometimes, disksize can be more than 100% But... I got ... /dev/ar0s1a: ... 0.5% fragmentation /dev/ar0s1e: ... 0.0% fragmentation /dev/ar0s1f: ... 0.0% fragmentation /dev/ar0s1d: ... 0.1% fragmentation Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1a248M 53M175M23%/ devfs 1.0K1.0K 0B 100%/dev /dev/ar0s1e248M -278K228M-0%/tmp /dev/ar0s1f221G1.4G202G 1%/usr /dev/ar0s1d248M 30M197M13%/var My questions: 1. How do I make /dev/ar0s1a 0.0% fragmentation the clean way? If I really wanted to? 2. How come /tmp is -0% in size? -278K? What had happened? as I have never experienced this in the previous installs on the exact same hardware. Thanks Lei ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation, 0%?
At 11:54 PM 8/13/2005, Lei Sun wrote: Hi, I know this question has been raised a lot of times, and most of people don't think it is necessary to defragment ufs, and from the previous posts, I got to know there are sometimes, disksize can be more than 100% But... I got ... /dev/ar0s1a: ... 0.5% fragmentation /dev/ar0s1e: ... 0.0% fragmentation /dev/ar0s1f: ... 0.0% fragmentation /dev/ar0s1d: ... 0.1% fragmentation Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1a248M 53M175M23%/ devfs 1.0K1.0K 0B 100%/dev /dev/ar0s1e248M -278K228M-0%/tmp /dev/ar0s1f221G1.4G202G 1%/usr /dev/ar0s1d248M 30M197M13%/var My questions: 1. How do I make /dev/ar0s1a 0.0% fragmentation the clean way? If I really wanted to? You don't. The term fragmentation does not mean the same thing in FreeBSD that it does in other OS's. (ie windows) Fragmentation in FreeBSD refers to blocks that have not been fully allocated. For example, if I have a file system that has 16K blocks, and 2K fragments (think of fragments as sub-blocks if it helps), and I save an 18K file, it will occupy 1 block and 1 fragment from the next block. That second block then is said to be fragmented. In the windows world, fragmentation refers to files which occupy non-contiguous groups of blocks. For example, you might have 5 blocks in a row, and then have to move to another part of the disk to read the next 5 blocks. 2. How come /tmp is -0% in size? -278K? What had happened? as I have never experienced this in the previous installs on the exact same hardware. Not sure about that one. Maybe someone else has an answer. -Glenn Thanks Lei ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation, 0%?
On Sun, Aug 14, 2005 at 12:09:19AM -0700, Glenn Dawson wrote: 2. How come /tmp is -0% in size? -278K? What had happened? as I have never experienced this in the previous installs on the exact same hardware. Not sure about that one. Maybe someone else has an answer. This is a FAQ. The available space is always computed after subtracting some space that would be only available to root (typically around 5% or 10% of the partition size). This free space is necessary to avoid internal fragmentation and to keep the file system going. Root may be able to borrow some space from this (in which case the capacity goes below 0%), but it is not advisable to keep the file system so full, so it should be only for a limited period of time. In your example, you're 278K over the limit; and should delete some files to make space ASAP. Should /tmp fill up more, it will soon become inoperable. -Glenn -cpghost. -- Cordula's Web. http://www.cordula.ws/ ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation, 0%?
At 12:18 PM 8/14/2005, cpghost wrote: On Sun, Aug 14, 2005 at 12:09:19AM -0700, Glenn Dawson wrote: 2. How come /tmp is -0% in size? -278K? What had happened? as I have never experienced this in the previous installs on the exact same hardware. Not sure about that one. Maybe someone else has an answer. This is a FAQ. The available space is always computed after subtracting some space that would be only available to root (typically around 5% or 10% of the partition size). The default is 8%. This free space is necessary to avoid internal fragmentation and to keep the file system going. Root may be able to borrow some space from this (in which case the capacity goes below 0%), but it is not advisable to keep the file system so full, so it should be only for a limited period of time. The reason for having the reserved space is to allow the functions that allocate space to be able to find contiguous free space. When the disk is nearly full it takes longer and longer to locate contiguous space, which can lead to performance problems. In your example, you're 278K over the limit; and should delete some files to make space ASAP. Should /tmp fill up more, it will soon become inoperable. From the original message: Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp This shows that /tmp is empty. If the reserved space was being encroached upon, it would show 100% capacity, and available bytes would go negative, not bytes used. It would look something like this: Filesystem SizeUsed Avail Capacity Mounted on /dev/ad0s1a248M238M-10M 105%/ I've never seen the capacity go negative before, which is why I suggested someone else might know the answer. -Glenn ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation, 0%?
Thanks for the good answers. But can anyone tell me why the capacity is going negative? and not full? Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp Thanks a lot Lei On 8/14/05, Glenn Dawson [EMAIL PROTECTED] wrote: At 12:18 PM 8/14/2005, cpghost wrote: On Sun, Aug 14, 2005 at 12:09:19AM -0700, Glenn Dawson wrote: 2. How come /tmp is -0% in size? -278K? What had happened? as I have never experienced this in the previous installs on the exact same hardware. Not sure about that one. Maybe someone else has an answer. This is a FAQ. The available space is always computed after subtracting some space that would be only available to root (typically around 5% or 10% of the partition size). The default is 8%. This free space is necessary to avoid internal fragmentation and to keep the file system going. Root may be able to borrow some space from this (in which case the capacity goes below 0%), but it is not advisable to keep the file system so full, so it should be only for a limited period of time. The reason for having the reserved space is to allow the functions that allocate space to be able to find contiguous free space. When the disk is nearly full it takes longer and longer to locate contiguous space, which can lead to performance problems. In your example, you're 278K over the limit; and should delete some files to make space ASAP. Should /tmp fill up more, it will soon become inoperable. From the original message: Filesystem SizeUsed Avail Capacity Mounted on /dev/ar0s1e248M -278K228M-0%/tmp This shows that /tmp is empty. If the reserved space was being encroached upon, it would show 100% capacity, and available bytes would go negative, not bytes used. It would look something like this: Filesystem SizeUsed Avail Capacity Mounted on /dev/ad0s1a248M238M-10M 105%/ I've never seen the capacity go negative before, which is why I suggested someone else might know the answer. -Glenn ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
disk fragmentation
during the boot sequence, I routinely see a % fragmentation message. It was my understanding that fragmentation doesn't occur on a Unix (er FreeBSD) box.. It seems that there is a concept of fragmentation from the above message, so, is there an un-fragment utility? Jim ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation
In the last episode (Feb 01), Jim Pazarena said: during the boot sequence, I routinely see a % fragmentation message. It was my understanding that fragmentation doesn't occur on a Unix (er FreeBSD) box.. It seems that there is a concept of fragmentation from the above message, so, is there an un-fragment utility? In the ffs filesystem, a file that's smaller than the default 16k blocksize (or the last part of a file that doesn't completely fit into a block) doesn't have to waste an entire block. Blocks can be split into eight 2k fragments and small files are put in them. The % fragmentation is just the percentage of fragment blocks vs the total number blocks. It's more an indicator of how many small files you have in the system than anything else. -- Dan Nelson [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: disk fragmentation
Jim Pazarena wrote: during the boot sequence, I routinely see a % fragmentation message. It was my understanding that fragmentation doesn't occur on a Unix (er FreeBSD) box.. It seems that there is a concept of fragmentation from the above message, so, is there an un-fragment utility? Jim No there is not a defragmenting program. Fragmentation is not a problem it is part of the normal operation of the filesystem. Data is stored in the filesystem in blocks, if a file does not have enough data to evenly fill all of its assigned blocks then the last block for the file is a fragment. The UFS filesystem will fill the fragment with new data when new data is added to the file. Some filesystems used by a redmond based company do not attempt to fill existing fragments, but simply add new blocks to the file so that a file could have more than one fragmented block. There are better descriptions of what is occurring in the archives, and presumably in textbooks that discuss filesystems. -- Jeremy Faulkner [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]