Subject: ANNOUNCE: mdadm 3.0-rc1 - A tool for managing Soft RAID under Linux I am pleased to announce the availability of mdadm version 3.0-rc1 It is available at the usual places: countrycode=xx. http://www.${countrycode}kernel.org/pub/linux/utils/raid/mdadm/ and via git at git://neil.brown.name/mdadm http://neil.brown.name/git?p=mdadm This is a "release candidate" which means that I think it is safe to use and that there will be no significant change in functionality before release. The man pages aren't really "release candidate" yet but I will be working on them before the final release. The most significant changes since -devel3 relate to the names of md devices as they appear in /dev and /dev/md/, and in particular the names that are used when an array is assembled with "--incremental" or with "mdadm --assemble --scan" when there are no ARRAY lines in mdadm.conf. In these cases mdadm needs to deduce a name to use, and to try to avoid using a name that a different array might have a stronger claim to. The rules are: - if the array is mentioned in mdadm.conf, use the name given there. - if the array appear to have been created for "this host" using the "homehost" concept, trust the name given in the metadata - if the new setting "HOMEHOST " is given (can be in mdadm.conf or on command line) the the name given in the metadata is not associated with some other array by mdadm.conf, then trust the name given in the metadata - otherwise use the name in the metadata, but in an untrusted manner. If a name is untrusted, or if the name is already in use by another array, then a numeric suffix like "_0", "_1" is appended to create a unique name for the array. That name is then used to create a device file in /dev/md/. So if all arrays needed for boot will always be listed in /etc/mdadm.conf, then it is appropriate to add "HOMEHOST " to mdadm.conf and there is no risk of conflicting names. However if you want auto-assemble to assemble all arrays at boot time and you don't want to list them in mdadm.conf, then don't give "HOMEHOST " either else there could be a risk of the wrong array being assembled for a given name. The following is the same introduction to 3.x as appeared in previous announcements. Any testing and feedback will be greatly appreciated. NeilBrown 11th May 2009 ===================================================== The significant change which justifies the new major version number is that mdadm can now handle metadata updates entirely in userspace. This allows mdadm to support metadata formats that the kernel knows nothing about. Currently two such metadata formats are supported: - DDF - The SNIA standard format - Intel Matrix - The metadata used by recent Intel ICH controlers. Also the approach to device names has changed significantly. If udev is installed on the system, mdadm will not create any devices in /dev. Rather it allows udev to manage those devices. For this to work as expected, the included udev rules file should be installed. If udev is not install, mdadm will still create devices and symlinks as required, and will also remove them when the array is stopped. mdadm now requires all devices which do not have a standard name (mdX or md_dX) to live in the directory /dev/md/. Names in this directory will always be created as symlinks back to the standard name in /dev. The man pages contain some information about the new externally managed metadata. However see below for a more condensed overview. Externally managed metadata introduces the concept of a 'container'. A container is a collection of (normally) physical devices which have a common set of metadata. A container is assembled as an md array, but is left 'inactive'. A container can contain one or more data arrays. These are composed from slices (partitions?) of various devices in the container. For example, a 5 devices DDF set can container a RAID1 using the first half of two devices, a RAID0 using the first half of the remain 3 devices, and a RAID5 over thte second half of all 5 devices. A container can be created with mdadm --create /dev/md0 -e ddf -n5 /dev/sd[abcde] or "-e imsm" to use the Intel Matrix Storage Manager. An array can be created within a container either by giving the container name and the only member: mdadm -C /dev/md1 --level raid1 -n 2 /dev/md0 or by listing the component devices mdadm -C /dev/md2 --level raid0 -n 3 /dev/sd[cde] To assemble a container, it is easiest just to pass each device in turn to mdadm -I for i in /dev/sd[abcde] do mdadm -I $i done This will assemble the container and the components. Alternately the container can be assembled explicitly mdadm -A /dev/md0 /dev/sd[abcde] Then the components can all be assembled with mdadm -I /dev/md0 For each container, mdadm will start a program called "mdmon" which will monitor the array and effect any metadata updates needed. The array is initially assembled readonly. It is up to "mdmon" to mark the metadata as 'dirty' and which the array to 'read-write'. The version 0.90 and 1.x metadata formats supported by previous versions for mdadm are still supported and the kernel still performs the same updates it use to. The new 'mdmon' approach is only used for newly introduced metadata types.