Introduction to the Ext4 File System for Linux
Вставка
- Опубліковано 20 тра 2024
- We take a look at the Ext4 file system for Linux.
Check out www.ezeelinux.com for more about Linux.
Please join the discussion at EzeeTalk.
www.ezeelinux.com/talk/ It's free, secure and fun! - Навчання та стиль
ERRATA: The default inode size for Ext4 is 256 bytes and not 256 killobits as I stated. The company that developed XFS for IRIX was SGI not CGI. I misread my own notes. The typo police have pointed out that it is "minix" and not "minux." :)
YAY! My request was made, I made it last year... Thanks Joe.
Thanks for that great explanation of the ext4 filesystem. When you mentioned you could theoretically have as many partitions as you like at 6:16 I actually tried it out. The result was:
Linux can handle up to 255 partitions per disk (e.g. "/dev/sda1" to "/dev/sda255"). If you have more than that, partitions beyond 255 won't show up in "/dev" (and can't be mounted). With gdisk it's no problem to create 255 (or even more) GUID partitions on a disk, but fdisk refuses to create more than 60 partitions with a DOS/MBR partition table. (Linux still recognizes up to 255 partitions if you manage to create them somehow else.)
21:39 The filesystem state should always be "clean" on a ext3 or ext4 filesystem even when it's mounted, but it says "needs_recovery" in the features section. That feature flag will be reset whenever the filesystem is unmounted properly. It's done this way for backward compatibility. The "filesystem state" will only change if the filesystem is mounted as ext2 (not using the journal at all) or if there is something wrong with the journal so "fsck" knows it can no longer rely on the journal but has to scan the whole filesystem to fix it. The other way around the "needs_recovery" flag prevents the filesystem from being mounted as ext2 (or checked by an old version of fsck that only supports ext2) to prevent breaking the journal. So only a properly unmounted ext3 (or ext4) filesystem can be mounted as ext2.
I paused at 0:45
To google "what is the common file system on linux"
now I know I should just watch the damn video lol
Great video and very informative to those of us wanting a little deeper dive.
In fact, there is no limit for the number of extended partitions besides memory in Windows. You can only use the drive letters up to z, but you can also, simular to Unix, mount a drive to a directory or using the subst command.
Fun fact: Most popular bootloaders happily accept use of logical partitions for boot. I've been doing that for years, which is especially handy to know for someone with a MBR table on primary disk with three parts - Modern open-source systems can exist in Extended just fine and operate entirely within that. Now GPT, that's really fun - You can have as many parts as you want for GPT, but the EFI partition can only have so many things in it as typically this is very size-limited. Fortunately, a separate EFI part on external media can be used, and is what I'd recommend.
Hey Joe! Thanks for another great video; love the content. Long time (mostly)power user from windows. 6 or 7 years ago I finally started to dabble in Linux after learning how to root Android phones. Went: ubu, mint, and Manjaro. So happy. I avoid win as much as possible now.
I'm super excited to see you mention butter in this video. It can be quite difficult to set up at first, but once you get a feel for subvolumes, it's quite nice. It is a little more resource intensive; my 2010/2011 Acer Aspire 7741G handles it pretty well though. I currently use 2x 2tb sshds in raid1 each drive with luks. Love the security from outside and to loss!
You brought it up, but I wanted to expand a little farther for your other lucky viewers: snapshots are directly linked to the (sub)volume being snapshotted. If magical gremlins corrupt a bit or two it CAN carry over into snapshots. You also touched on timeshift and you're absolutely correct! Btrfs makes my timeshift backups instantaneous! Literally! Also you can mount snapshots live and pull files out, without stopping anything else.
Another great plus is btrfs' ability to immediately accept new drives into a pool and balance data live. You can also convert between jbod and raid{0,1,5,6,10) live, and in the background, as well.
Finally, and maybe the most intriguing to me, is you can actually use btrfs on a whole disk. A partitionless btrfs drive. No msdos or gpt. I'm not using it for my system, but I find it interesting that you don't even need a partition table, just a giant lump of btrfs. :)
Thanks again for the video! Always love the content! Hope you and your fam are staying safe through everything!
thanks Joe it helped me understand a bit more about files systems
Hey, good Video! And as a Feedback i am an Former Windowsuser and i was really amazed how fast ext4 handles files also. I did copy some Gigabytes of photos and i was really impressed how fast it did that. In my NTFS it would have took so much longer!
One of the edge cases for running out of inodes is when you're running a Usenet news server or mail server back in the 90s at least. It was common to run out of inodes given all the small message files that were often less than 4K or even less than 1K in size. One of the fixes for that was to change the bytes per inode size to 1024. There were also special filesystems or raw storage files that you could use to get around this problem. It's been 20 years since I ran a usenet server, but generally these days the file systems are so big that you never run out. Still, you can check this with df -i
Joe talks about file systems? Let me hit that thumbs up button and get a coffee, I am in.
hi , should i accept that efi is required for installing debian based one ? can install linux without this efi thing ??
@@pika9985 you can refuse by simply useing legacy mode(csm) somewhere in the bootoptions. I still dont see the harm in installing your debian with (u)efi enabled.
@@fernabianer1898 would you tell me what to flag on root partition (flag as boot not root?) to bot directly from it , not from uefi small fat32 thing ? how could i do ? thanks .
@@pika9985 no need for that. set your bios to csm/legacy ONLY. insert your boot medium, install. thats it.
@@fernabianer1898 thank you for replay , i use application called "balenaEtcher" to write the img on usb medium , it make the usb has simple user interface "small partition.." if it nothing wrong with that i go for your tip < sorry if my question seems stupid .i'm beginner
This is really informative, thank you so much for making this video.
How to run out of inodes:
Step 1) Have a large, embarrassingly parallel, problem you want to solve
Step 2) Don't appreciate power of combinatorics
Step 3) Your system is out of inodes.
This is good stuff. Thank you.
I've seen exhausted inodes one time in my entire life. That was a development server with countless docker containers and it's volumes mounted in the root file system. Pretty funny situation when you have 50 percent of free space but you can't write anything and ends with read-only file system.
How do docker filesystem files which are usually big exhaust inodes? The classical inode exhaustion happens on Mail or News servers with typical files smaller than 4K.
@@berndeckenfels development applications can be messy and are constantly abused by QA. So when you have a dev server that nobody logged into last year because CI pipelines did all the job for dozens of projects then it is no problem for docker volumes to exhaust the inodes. Note that I haven't said it's common to docker to do that. It just happened to me.
I'm trying to get into BSD kernel development, and this really helped.
This is the content I want to see!
You can have snapshots with LVM too, without ZFS or BTRFS. Contrary to the video, I suggest using LVM, even if you don't need to, now. Later it could save you time, for example when you want to take a snapshot, want to add storage, want to rearrange the partitioning without actually have to repartition the drive. LVM is easy.
Every time Joe said Msdos, I was having flashbacks to programming in 1978...
very well explained, thank you for that
Very informative and helpful
Thank you for this video!
I recently started using BtrFS on a laptop when I replaced the HDD with an SDD. It runs fast and the snapshots are a great feature that makes it worthwhile even on a single disk with one partition.
Otherwise I generally agree with your assessment - ext4 is the way to go for basic systems. My main system uses ext4 on an SSD for boot/root and an LVM2 RAID1 HDD array with ext4 for the home partition. This box is now old - six years old - and still runs fast and reliably though I've had two HDD failures. I didn't even notice that a drive had died until the system told me via e-mail. Replace the dud disk, copy the partition scheme to the new blank disk, add the new disk to the array and let the array rebuild while doing other things. So reliable that it's boring - which is a good thing. :)
I'd say JFS and NTFS are only for use with disks that are already formatted that way to which you need access. Mount them, get the data off and get them out of there. I'd never leave an NTFS disk in a Linux box for regular access. Use a Samba share for Windows compatibility.
ZFS is great for multi-disk (RAID) storage solutions but I'd be inclined to use it only with a BSD based system such as FreeNAS. Less fiddling to get it set up.
XFS I think is a corner case useful only for systems which have those particular demands which is why SGI built it that way in the first place. Nice to have available to Linux if you find you have a disk with files on it that you need to access but not practical for most general applications. Your database example is also an appropriate application.
Shouldn't that be 'Fetid Dingo Kidneys"?
Yeah but it's NOT ... Better Than Reiser File System,,,same thing
Number of indies can be set at filesystem creation time. Inifes cost space, so on a filesystem for let’s say database files where there will only be a small number of files, it is common to reduce the qty, and thus gain available space.
Thanks for the good information!
great video! , clear explanation, I'm searching for this and your video is the best
Awesome stuff!
Great overview video oder Ext, but there are three things I'd like to mention:
- You should limit the inode count on very big data disks with mkfs.ext4 -i or -N. The default creates one inode per 16kiB which results in a loss of 1/64 of the disk space which can be a lot. I recommend 1-2 million inodes per TB. If you're unsure, you can go 5 million and still save a lot of space.
- You can increase the number of inodes by and ONLY by increasing the size of the partition. That's because there is a fixed number of inodes on block-group-basis. One block group will usually hold 32768 blocks or 128 MiB.
- NTFS isn't AS bad as you described in the end of the video. I was running all my data disks (multiple TiBs) on my Linux desktop for almost 2 years without issues until recently. The only thing with NTFS is that it's unbearably slow on Linux, that's why I'm currently migrating to ext4.
Thank You Joe!!!
very informative and helpful. thanks!
btrfs also supports transparent compression - decompress is auto, but compress is a amount option. ZedFS is the only other Linux file system with compression.
You can still update files in place (that's why you can update without restarting). If you remove a file that is in use you will only remove the link the parent directory of the file has to it. The file (and inode) will remain on their places until all the links are removed (when the file is no longer being used).
I wish more people would appreciate how awesome UEFI and GPT technology is. Dual booting and your partition layout is MUCH MORE versatile that way. Being dumped into legacy grub really sucks, with a UEFI system you can be in windows 10, shift click on reboot, and select your linux from the windows boot manager and boot into it, what's more cool than that? What's more cool then pressing your boot option key at boot and being able to select windows 10 or your linux without even seeing grub (until you select linux)...it's freaking cool guys trust me.
super informative video . Thanks
Not sure if this has been addressed or if you found out after the fact already, but the Magic Number you saw in terminal is the hex identifier for ext4 (I think)! Basically, if you looked at the boot sector for an ext4 partition with a hex viewer tool you would see EF53.
All filesystems have some kind of hex identifier though I can't recall any others off the top of my head.
It can be usefull to know the inodnum (ls -i) on a file with a name you can't type and is a hell to delete. Then one can use find . -xdev -inum xxxxxx -ok rm {} \;
I like btrfs since it's so easy to mirror and even allow hot replace of a mounted disk.
It's not slow unless you write more data than there is free cache memory, but it's not stable enough yet, it's more stable on suse than on mint but it's hard to fix if it crash, so it's not for beginners.
again a good class
Nice video, learned something new today.
I know you were joking about a 16 TiB file but I think we aren't so far away from it, videos or video editing can create massive files, especially the film industry. I for myself have "only" a 4.3 GiB large photo file.
If you were editing a film, you'd be better off with XFS. That's exactly what it was designed to do. :)
Thanks Joe for the video!
I always learn something, today I learned to stick with Ext4 and my old eBay junk will run just fine...
I don't buy new stuff, I reuse old stuff from eBay...
I am not a Gamer or a power user!
I just watch Movies and listen to Music. I record and edit both of them with my 4 core 16 gig systems and they do what I need them to do... \ : - )
Simple life for a simple man... Going on over 2 years on Linux, thanks to you Joe.... Thank You!
LLAP
If you have a file system for only large vide files, you should consider xfs, as it is designed for handling/streamimg large files.
Extfs is not...
Yes spot on Glazed over sounds about it
While I would agree with you that Ext4 is the default file system in most common Linux distributions, you are incorrect in saying that Ext4 is the default file system for Arch Linux. It's certainly the default file system for many distributions based on Arch, but it's not the default file system for plain vanilla Arch. Arch Linux has no default file system because the person installing Arch Linux is responsible for explicitly choosing a file system to create on the disk that they will be installing Arch onto. There is no menu of choices with a default choice pre-selected.
While I agree, as an Arch Linux user I've noticed that much of the documentation sort of assumes you're running ext4. The Wiki will often briefly mention alternative steps for other file systems but ext4 is considered "the usual one", even on the wiki. So yeah, I wouldn't say he's "incorrect".
Did you know that ext3 and ext4 filesystems can be grown and shrunk on a live (i.e. running) system? See the resize2fs command. Works well with LVM.
ZFS is the default filesystem for some BSD systems (e.g. FreeBSD), so its definitive source is quite stable. Its port to Linux might be a different story, which may be why it's still considered experimental.
Shrinking should only be possible on unmounted file systems afaik.
I use Fedora MATE mostly and it still defaults to Ext4. It defaults to LVM, which I change to standard partitions, but it's auto-partition creates a separate /home partition, which I don't like personally, so have to delete it and max out the / partition size. I've seen some demos of btrfs and it seems kinda interesting, but nothing I think I need.
For the NTFS portion of the video, I'd like to point out that the native Windows file system would only really slow down the computer over time if the OS is installed on a hard drive disk, otherwise, with SSDs, it wouldn't be likely to happen. Another addendum, I don't think having an NTFS partition on Linux would be a problem if it's not storing the operating system nor related files, but, for instance, Windows (in this case, Linux would really just read the data, right?)
At 28:40 you seemed to imply that a file could have more than one inode if it had more than one link to it. This isn't quite correct; a file can have as many links as necessary but all have the same inode. What the ln command does (at least for hard links) is make a directory entry with a new name if in the same directory or any name if in a different directory but each of them references the same inode. The link count within the inode keeps track of how many directory entries reference a file. This means that the rm command actually removes a file and frees up the inode and associated data blocks by deleting the last directory entry and decrementing the link count to zero then using the unlink system call to free the inode and data blocks. The ln command to create a soft link is a completely different kettle of fish as a soft link is a separate file containing a reference to a file anywhere in the directory tree. The (hard) links described above are restricted to a single file system.
The inode numbers help if you are looking for hardlinks
I've been using BTRFS since 2010, it is ok to use. I have not lost data for using BTRFS.
When I started it was ext2. Never kept up with what the changes were. Didn't worry about it much because it works. Now to catch up.
Biggest change was Ext3, with it's journal. Ext4 it mostly a progressive update (higher size limits etc.) but I guess the biggest change was the use of extents, for a more efficient allocation scheme on the disk.
As of the release of OpenSuse Leap (15.0) in 2015, brtfs became the default for Suse. It was still an option before this, but not default.
awesome video thanks, slightly related question, I'm just wondering why the entire / directory would be moved into lost+found during an fsck, could that be an unlucky event where the inode table was corrupted or something else?
I've never seen that happen. If it does, just reformat the drive and start over.
O ode numbers are useful for finding what names are linked to a file.
Some comments are also interesting here. Thank you.
Defragmenting SSD is needed if you want to lower the size of the partition. SSD life is so long now, with auto protection against failure. So a few defrags don't shorten it's life.
Microsoft refuses to undo it's legal hold on NTFS. The open source version of NTFS is not as good as the latest version of the Microsoft version.
Later versions of living partitions still are being developed have extra features. Self repairs in any of several undocumented ways is one secret feature.
On non business computers, the inbuilt compression allows more storage in limited space. It might also allow faster access times.
I am using ZFS on Linux, as rootfs, and even with 64GB of RAM on a dual Xeon server it is painfully slow under lot-of-small-files load. I mean, a deeper apt upgrade or copying a large source code tree could grind the system to halt. I am yet to understand how to tune it properly. 🤔
Would be way nicer than RAID+LVM+Ext4...
Jfs are for large files fast, like xfs.
Zfs is for servers, like btrfs. Btrfs was made as a Linux copy of Zfs. But Zfs is made by Sun for large servers. Btrfs still miss lots of the features in Zfs.
Hmm, 16TB file size limit is becoming an issue with that large hard drives available. That's with 4kb blocks though, so you can make the blocks larger ( up to 64kb limit for some platforms ) to go around this limit.
Do you know of any software that will display the file system like the first graphic, with the root at the top and the folder/sub folders underneath ? That would be a great tool to study a file structure on a machine.
Yes, it would! Some file managers can show a tree on the left hand side but which ones do that I'm not too sure.
There are the tree command, I guess.
And there are also som graphical tools I only use one or two times a year, so I don't remember their names. 😜
I've actually run out of inodes on an email server where each message is stored as a single file. When you get into multiple thousands of boxes and messages per box. Well, you get it.
I have reiserfs to my cyrus, it has no inodes, it just fake them.
@@ytbabbler this was probably 15 years ago, nowadays it's strictly COW file systems, everyone is delete happy and it saves a lot of trouble if you take regular snapshots.
@@booleanenator My cyrus server is older than that so it still has reiserfs.
But since there is one file / email and those files never changes, it's no need for snapshots, it's very easy to do a normal backup that can be run many times per day. COW is completely irrelevant when files are created once and never change.
@@ytbabbler I use it in an array, so COW solves the RAID-5 write hole. I am not worried about files that don't change, and to be clear. That e-mail server was years ago. I am not in the ISP/Telecom business anymore. COW has more than one advantage.
@@booleanenator Yeah, but my comment was only for the cyrus pools, I found that reiserfs was good for this but it's the only place I still use it. I use xfs for video editing and btrfs for some stuff and testing but ext4 as default.
I was working with advfs and zfs for many years and like that stuff too but for linux it will probably be openzfs due to license problems with zfs.
I had a nas with freebsd/zfs but there was not enough with memory so it didn't perform well.
My take on filesystems is that there is not one that is best on everything, they have different features and strengths/weaknesses.
Thanks
I almost always wondered about this Ext file system on Linux...
@Peter Mortensen The way it works or the way they categorize the file systems...
@@SkyFly19853 all file systems in Unix/Linux has the same abstraction around inodes etc. And those are constructed by Linux if a file system doesn't have those, like NTFS and FAT.
@@AndersJackson
I see.
"You can see extra numbers next to the time stamps". What extra numbers? I can't see them. (At 21:59)
‘[…] the time stamps are actually extremely accurate’ … I assume he talks about … seconds?
45:55 no, SUSE uses btrFS by default, not XFS
No, they use both. Btrfs on / and XFS for /home.
openSUSE definitely defaults to BTRFS, not XFS. Watch out, because with its snapshots it ran through the root partition of my old laptop like candy, I had to disable them. When I moved it to an SSD, I completely reinstalled on ext4 it just to get rid of BTRFS. I think SLES uses XFS, not sure. RHEL yes, now it defaults to XFS. I also installed a Fedora WS not that long ago, I don't really remember but I would say it was ext4 by default.
Knowing how directory files and inodes work can be used to 'undelete' a file assuming the file has not been overwritten.
MBR is 4 primary partitions maximum
Good dam class
I use to run defrag after installing xp
I find it strange, that MS stopped developing their new filesystem (can't remember the name it had now). And decided to keep NTFS. Didn't they get it to work?
Well, anyway. Running BTRFS on my main rig with Garuda. Dual booting with W10. Though choosing which to boot with uefi boot selector. Not traditional "dual booting"...
At 41:45 you say that inodes are 256 kbits. Is that right? Sounds huge if there is one for each file! Did you mean 256 bits?
Glad you caught that... Yeah, I misspoke. It's 256 bytes. Thanks. :)
Hey what's your opinion on f2fs. Android Phones seem to ship with it.
I don't know enough about it to have an opinion.
Strangely I've come across situations on external HDDs with NTFS where Windows "tries" to fix it but then tells me it's unfixable ... do I want to format it. I then plug it into something like Ubuntu, all the files and folders are there, I can copy them off, etc. What was even more surprising, if I then pluged the drive back into Windows ... lo and behold, Windows now thinks it's fixed, no more issues.
But you are correct, the ntfsfix utility is not as comprehensive as the Windows chkdsk tool, it may not be able to fix every possible corruption. Personally I prefer including the optional ntfsprog package so the normal fsck works on NTFS as well.
Interesting.... Makes you wonder what's going on withe drives. :)
@@EzeeLinux Well it was on externals, so my guess was that someone unplugged it without "safe remove". Perhaps something like the MFT space was corrupted due to an incomplete write (that could seriously damage a NTFS drive).
What is worrisome is how come did they get fixed by just plugging it into Linux? Is MS's NTFS driver so buggy that could not handle a corruption which was easily fixed? Possibly not surprising if comparing the recently open sourced ExFAT source code with the older ExFAT from Samsung.
@@benriful or...
Linux didn't bothered about some cases that could mean loosing data and fixed it, ignoring that.
The proper thing would probably be to make a backup, reformat and then restore the backup.
You do have backup, don't you?
All data not backed up is just cache and not important. 😜
@@AndersJackson Agreed about the backup. This was to "fix" someone else's disks though. The entire point was to get the files off it. And that I achieved by simply plugging it into a Linux and copying off.
What was strange to me was that after this, Windows found those drives to be fine. Even with a chkdsk it found no errors.
I did advise that the drives would need a format at least, and that they "might" be at their end of life. These were rather old.
Hey Joe. Do i need swap partition to install the new Ubuntu 19.10?
No, it will create its own swap file. :)
@@EzeeLinux what do you think, 32 gb SSD is enough to run Ubuntu 19.10?
Windows Uses the LRU Algorithm for storing memory to disk ... Right?
how did you manage to modify a file precisely at a time with all 0s after the dot? (from 28:10 on)
I would bet that it's because the system is using relative times because of the fact that it's running an SSD. It's a setting that helps reduce write too disk. :)
@@EzeeLinux Sorry, I still don't get it. Relative to what? How does that relate to it being an SSD? Is it explained in some other video?
@@nicolasfiore Early SSD's had a linted number of write cycles before they'd stop working properly. SSD's still need to reduce the number of writes to last a long time but it's much better now. Relative access times are a way to reduce writes to a file because the times are not written every single time the system access that file. It only marks it once per day to indicate that the file was access or modified. The idea is that it reduces writes and saves SSD life. The setting is automatically turned on for the file system when the installer detects an SSD.
The comment on ntfs breaking itself all the time is only partially right IMHO. An ntfs system flags itself as "dirty" easily and if you don't umount cleanly e.g. pull the usb cable a bit too quickly, it will not clear the dirty flag, while there is likely nothing wrong with the filesystem (or something "trivial"). There is a tool to fix some common ntfs problems from linux (but not all) - ntfsfix. Ntfsfix can clear the dirty flag, and if the filesystem is more seriously damaged - it will schedule it for check next time it runs in windows.
What about the bad block table and marking bad blocks and failing blocks?
That's in there. Can't do it all in a video. :)
I realize this is extremely unlikely since writing a line of text takes only a few cycles, but just as a thought experiment, what would happen if your computer lost power while writing to the journal itself? Would fsck revert to the slow ext2 grind, since it realizes the journal is corrupted?
I'm not sure... The journal is structured in such a way as to be rather robust. While it can be corrupted, I think it's not something that is very common under normal circumstances. I do know that Ext4 recovers very well from a power oll and I've never had an Ext4 system that corrupted itself beyond simple repair. :)
@@EzeeLinux I do not know for sure, but there are probably some checksum mechanisms at work to keep track until which point the writing of the journal is considered clean.
if u have 26+ partitions u will get sda27 and if u have more drives u will get sdz, sdaa, sdab...
Now make a video on ZFS.
thank you very much joe learned a lot : )
What do you think, 32 gb SSD is enough to run Ubuntu 19.10?
It will run fine in that space but it doesn't leave much room for adding data file... Depends on what you wan to do I guess... :)
What happens when INodes start growing into data block territory ? Do the subsequent INODES move to the next free space leaving behind a "continued.." link ?
if data allocation happens from the other end then i can see it'll be a non issue
Does it actually go like this though ?
@@animatrix1851 The number of possible inodes is fixed when the file system is created. It allocates enough space for the table. The number of inodes can't be increased. You'd have to reformat the disk with a bigger partition to do that. :)
@@EzeeLinux ah that makes more sense thanks! Could you make a similar video explaining f2fs ?
Also, you'd gone over a situation where you could theoretically have more inodrs than files. This wouldn't be possible with a more normal formatting plan right ?
@@animatrix1851 I think you meant more files than inodes, right? Yes, it is possible to run out of inodes, especially if you're trying to put a lot of very small files onto a small partition. That's very rare though. You get plenty of allocated inodes for even the smallest drive. :)
tell me what is your microphone
Can you please talk about kubernetes and docker
No, this channel is about desktop Linux.
can you insert a kernel into the file system.
What about getting past the file system into the structure of the hard drive control mechanism I hear seagate are sending data from there hard drives over the internet.
I'd like somebody to get into the actual structure programming of the hard drive not just a file systems the actual operational processes behind the scenes that's where you want to stick your kernel. You shouldn't defragment SSD anyway shortens it's life and of course it's not quicker either ridiculous maybe thousands of a microseconds
How would seagate be able to control a Net device from a disk? Just ridicoulus until you make a demonstration.
And Yes, it is possible to monitor everything that goes out of your router, so it would turn up there.
File systems, raids and LVM2 do look at the underlying storage device capacity and settings to perform well. It has to.
Actually, you get some performance boost from defragmenting SSD diska too. But it is usually not worth it. It will show on huuuge data transfers, like movies.
But probably not worth it.
And guess what, the kernel IS stored in a file system. So your first question doesn't computer for me.
The max number of inodes is in the region of 4 billion (4294967296). If you assume 32bit inode. I don't think I have that many files :) .
Number of inodes are fixed, so after creating the file system, you can't easily change that.
You decide when the file system is created, and most people just use default values.
I use xfs on my 2gb ram Intel celeron laptop to make it run faster as compared to ext4
ntfsfix? Can it not fix NTFS file systems???
Not as well as old chkdsk. It can diagnose some issues but it's best to let Windows take care of NTFS if you can. :)
24 000 views. Interesting)
Mature file systems are the ones to use. Try to explain to your boss why your servers use JAZZFS128 after a major disk corruption.
Minix, not Minux.
Shit happens...
This video is astonishingly unclear. So much more potential for visual explanations with diagrams
The video is full of diagrams... What more would you have liked to see?
There is no Minux there is Minix!
There's no way ZFS is considered experimental. Maybe experimental on Linux through FUSE, but generally it's considered to be "The only filesystem you should ever trust your data on" (so long as you don't use filesystem compression inappropriately)
It is most certainly experimental on Ubuntu. Listen closer for those context clues...
Don't make a swap partition, make swap files.
Don't make swap.
@@mina86 Sometimes yeah.
@@rautamiekka, nope, never.
On any modern personal computer you’ll never run out of RAM.
On a workstation for memory-intensive tasks such as high-definition video editing, once you start hitting swap, your machine will be too slow to be productive so just buy more RAM. On top of that, big files will me mmaped so the system will be able to ‘swap’ them out even without a swap.
On a server, set up alerting to warn you if you ever start getting close to no memory and then buy a new unit.
@@mina86 Whatever.
Hibernate works better with a partition. So: No.