Linux Today: Linux News On Internet Time.

Red Hat: ext3 information

Oct 11, 2001, 05:41 (178 Talkback[s])
(Other stories by Michael Johnson)
Subject: ext3 information
From: Michael K. Johnson 
Date: Tue, 21 Aug 2001 17:53:08 -0400

I wrote up a short piece that I hope to flesh out a bit more later
on why Red Hat chose to include ext3 in this release, why you want
to use it, and what we did to make it robust.

It's not an anti-any-other-filesystem tirade at all.  Don't take
any part of it as meant to put down any other filesystem, even
ones we have not chosen to ship yet.  No hidden agenda involving
alien abductions...  :-)

Anyway, I hope it's useful.  Feedback to the list, please.


 "He that composes himself is wiser than he that composes a book."
 Linux Application Development                     -- Ben Franklin

Why do you want to migrate from ext2 to ext3? Four main reasons: availability, data integrity, speed, and easy transition.

After an unclean system shutdown (unexpected power failure, system crash), each ext2 file system cannot be mounted until its consistency has been checked by the e2fsck program. The amount of time that the e2fsck program takes is determined primarily by the size of the file system, and for today's relatively large (many tens of gigabytes) file systems, this takes a long time. Also, the more files you have on the file system, the longer the consistency check takes. File systems several hundreds of gigabytes in size may take an hour or more to check. This severely limits availability.

By contrast, ext3 does not require a file system check even after an unclean system shutdown, except for certain rare hardware failure cases (e.g. hard drive failures), because the data is written to disk in such a way that the file system is always consistent. The time to recover an ext3 file system after an unclean system shutdown does not depend on the size of the file system or the number of files; rather, it depends on the size of the "journal" used to maintain consistency. The default journal size takes about a second to recover (depends on the speed of the hardware).

Data integrity:
Using the ext3 file system can provide stronger guarantees about data integrity in case of an unclean system shutdown. You have a choice of how carefully to protect your data. Essentially, you can choose either to keep the file system consistent but allow for damage to data on the file system in the case of unclean system shutdown (for a modest speed up under some but not all circumstances) or to ensure that the data is consistent with the state of the file system (which means that you will never see garbage data in recently-written files after a crash.) The more safe choice to keep the data consistent with the state of the file system is the default.

Despite writing some data more than once, ext3 is often faster (higher throughput) than ext2 because ext3's journaling optimizes hard drive head motion. You can choose from three journaling modes to optimize speed, optionally choosing to trade off some data integrity. One mode, data=writeback, limits the data integrity guarantees, allowing old data to show up in files after a crash, for a potential increase in speed under some circumstances. This mode, which is the default journaling mode for most journaling file systems, essentially provides the more limited data integrity guarantees of the ext2 file system and merely avoids the long file system check at boot time. The second mode, data=ordered (the default mode), guarantees that the data is consistent with the file system: recently-written files will never show up with garbage contents after a crash. The last mode, data=journal, requires a larger journal for reasonable speed in most cases and therefore takes longer to recover in case of unclean shutdown, but is sometimes faster for certain database operations. The default mode is recommended for all general-purpose computing needs.

Easy transition:
It is easy to change from ext2 to ext3 and gain the benefits of a robust journaling file system, without reformatting. That's right, no need to do a long, tedious, and error-prone backup, reformat, restore operation in order to experience the advantages of ext3. There are two ways to do the transition:

  • The Red Hat Linux installer program will offer to transition your file systems when you upgrade your system. All you have to do is check one checkbox per file system.
  • The tune2fs program can add a journal to an existing ext2 file system. If the file system is already mounted when it is being transitioned, the journal will be visible as the file ".journal" in the root directory of the file system. If the file system is not mounted, the journal will be hidden and will not appear in the file system. Just run tune2fs -j /dev/hda1 (or whatever device holds the file system you are transitioning) and change "ext2" to "ext3" on the matching lines in /etc/fstab. If you are transitioning your root file system, you will have to use an initrd to boot; run the "mkinitrd" program as described in the manual and make sure that your lilo or grub configuration loads the initrd. (If you fail to make that change, the system will still boot, but the root file system will be mounted as ext2 instead of ext3 -- you can tell this by looking at the output of the command "cat /proc/mounts") More information on tune2fs can be found in the tune2fs man page.

A list of reasons Red Hat chose ext3 for our first supported journaling file system follows. Note that these reasons are not necessarily each unique to ext3 (some other journaling file systems share several of the points here) but the whole set of reasons taken together is unique to ext3.

  • ext3 is forwards and backwards compatible with ext2, allowing users to keep existing file systems while very simply adding journaling capability. Any user who wishes to un-journal a file system can do so easily. (Not that we expect many to do so...) Furthermore, an ext3 file system can be mounted as ext2 without even removing the journal, as long as a recent version of e2fsprogs (such as the one shipped in this release) is installed.
  • ext3 benefits from the long history of fixes and enhancements to the ext2 file system, and will continue to do so. This means that ext3 shares ext2's well-known robustness, but also that new features are added to ext2, they can be carried over to ext3 with little difficulty. When, for example, extended attributes or HTrees are added to ext2, it will be relatively easy to add them to ext3. (The extended attributes feature will enable things like access control lists; HTrees make directory operations extremely fast and highly scalable to very large directories.)
  • ext3, like ext2, has a multi-vendor team of developers who develop it and understand it well; its development does not depend on any one person or organisation.
  • ext3 provides and makes use of a generic journaling layer (jbd) which can be used in other contexts, and can journal not only within the file system, but also to other devices, so as NVRAM devices become available and supported under Linux, ext3 will be able to support them.
  • ext3 has multiple journaling modes. It can journal all file data and metadata (data=journal), or it can journal metadata but not data (data=ordered or data=writeback). When not journaling file data, you can choose whether to write file system data before metadata (data=ordered; causes all metadata to point to valid data) or not handle file data specially at all (data=writeback; file system will be consistent, but old data may appear in files after an unclean system shutdown). This gives the administrator the power to make the trade off between speed and file data consistency, and to tune speed for specialized usage patterns.
  • ext3 has broad cross-platform compatibility, working on 32 and 64 bit architectures, and on both little-endian and big-endian systems. Any system (currently including many Unix clones and variants, BeOS, and Windows) capable of accessing files on an ext2 file system will also be able to access files on an ext3 file system.
  • ext3 does not require extensive core kernel changes and requires no new system calls, thus presenting Linus no challenges to integrating ext3 into his official Linux kernel releases; ext3 is already integrated into Alan Cox's -ac kernels, slated for migration to Linus's official kernel soon.
  • The e2fsck file system recovery program has a long and proven track record of successful data recovery when software or hardware faults corrupt a file system. ext3 uses this same e2fsck code for salvaging the file system after such corruption so it has the same robustness against catastrophic data loss as ext2 in the presence of data-corruption faults.

Again, we don't claim that every one of these points are unique to ext3. Most of them are shared by at least one other filesystem. We merely claim that the set of all of them together is true only for ext3.

Here are some of the things Red Hat has done to ensure that ext3 is safe for users to use for their data:

  • We have done extensive stress testing under a large set of configurations. This has involved many thousands of hours of "contrived" load testing on a wide variety of hardware and file system configurations, as well as many use case tests.
  • We have audited ext3 for multiple conditions, including memory allocation errors happening at any point. We have tested that by forcing false errors and testing file system consistency.
  • We audited and tested ext3 for poor interactions with the VM subsystem, finding and fixing several interactions. A journaling file system puts more stress on the VM subsystem, and we found and fixed bugs both in ext3 and in the VM subsystem in the process of this audit and these tests. After thousands of hours of this testing, we are extremely confident in the robustness of the ext3 file system.
  • We have done an extensive year-long-plus beta program, starting with ext3 on the 2.2 kernel series, and then moving forwards to the 2.4 kernel series. Even before the official beta program, ext3 was put into production use in some circumstances; ext3 has been in production use on some widely-accessed servers, including the rpmfind.net servers, for over two years.

Related Stories: