xlv_mgr(1M) xlv_mgr(1M) NAME xlv_mgr - administers XLV logical volume objects and their disk labels SYNOPSIS xlv_mgr [ -r root ] [ -c cmdstring ] [ -v ] [ -x ] [ -R ] DESCRIPTION xlv_mgr displays and modifies existing XLV objects (volumes, plexes, volume elements, and XLV disk labels). xlv_mgr can operate on XLV volumes even while they are mounted and in use. xlv_mgr supports the following command line options: -r root Use root as the root directory. This is used in the miniroot when / is mounted as /root. -c cmdstring Command line option. Causes xlv_mgr to go into command line mode and use cmdstring as the command input. Multiple instances of -c are accumulated. -v Verbose option. Causes xlv_mgr to generate more detailed output. -x Expert mode. Provides additional functions. -R Do not read the XLV configuration information from the disks at the time xlv_mgr is invoked. This allows xlv_mgr to quickly startup. The XLV configuration information can be read from the disks at a later time with the reset command. Commands that pertain to plexes are displayed only when the system has been licensed for the plexing portion of XLV. xlv_mgr provides several types of operations: attach, detach, delete, change, script, and show: attach Add an XLV object to another XLV object. You can add a volume element to a plex or plex to a volume. The volume element or plex to be added must first be created using xlv_make(1M). detach Separate a part of an XLV object and make it an independent (standalone) XLV object. For example, if you detach a plex from a plexed volume, that plex is separated from the volume and made into a standalone plex. The original volume then has one less plex. delete Delete an entire XLV object. script Generate the xlv_make commands required to create some or all XLV objects. show Display the list of XLV objects on the system and their structure. change Change an attribute associated with can XLV object. The commands are: show [-long][-verbose] all Displays all known XLV objects by name and type. The -long option causes more information about each XLV object to be displayed. The -verbose displays more detailed information, such as the uuid. The following is an example of the output of this command: xlv_mgr> show all Volume: root_vol (complete) Volume: db1 (complete) Volume Element: ve12 Plex: plex2 show [-short][-verbose] kernel [volume_name] Displays all XLV objects or the specified volume_name configured into the running kernel. The only XLV objects in the kernel are volumes. Standalone plexes and volume elements are not viable objects in the kernel because they cannot be used. The -short option displays terse information on each volume. The -verbose displays more detailed information on each volume. show [-long][-verbose] labels [device_volume_header] Displays XLV disk labels on all disks or the specified device_volume_header. The -long option displays the secondary label. show config Displays XLV software configuration information about the running kernel. Included are statistics on the XLV memory pools for all plexed volumes (if support is present) and all volume elements. For example: xlv_mgr> show config Allocated subvol locks: 30 locks in use: 8 Plexing license: present Plexing support: present Statistic Collection: off Plex buffer pool statistics: maximum concurrent I/O's 13; I/O buffer size 1016 bytes buffer pool size 13.21KB; maximum size 4.00GB pool hits 1121; misses 141; waits 0; resized 4 times memory growth rate 50%; maximum miss rate 10% Volume Element buffer pool statistics: maximum concurrent I/O's 53; I/O buffer size 1.48KB buffer pool size 80.32KB; maximum size 4.00GB pool hits 1636; misses 225; waits 0; resized 5 times memory growth rate 50%; maximum miss rate 10% Maximum subvol block number: 0x7fffffff The Statistic Collection tag indicates if statistics gathering is enabled (on) or disabled (off). In the pool statistics, hits imply an efficient I/O path and misses suggest a somewhat slower path. When the maximum miss rate is exceeded, the buffer pool is grown by the growth rate. The buffer pool size is limited to maximum size. The maximum concurrent I/O's gives an upper bound on the number of efficient I/Os possible, and the I/O buffer size is the amount of memory dedicated for each I/O request. show [-verbose] object object_name Displays detailed information on an XLV object object_name. The information includes all the XLV parameters and the disk partitions that make up the object. In the example below, the volume named db1 has one subvolume of type data that contains two plexes. The first plex has two volume elements, while the second plex only has one volume element. The first volume element in each plex covers the same range of disk blocks. For each volume element, xlv_mgr displays the partitions that make up the volume element, the size of the partition, and the range of this volume's disk blocks that map to the volume element. For example: xlv_mgr> show object db1 VOL db1 (complete) VE db1.data.0.0 [active] start=0, end=1100799, (cat)grp_size=1 /dev/dsk/dks1d4s0 (1100800 blks) VE db1.data.0.1 [active] start=1100800, end=2201599, (cat)grp_size=1 /dev/dsk/dks1d4s1 (1100800 blks) VE db1.data.1.0 [active] start=0, end=1100799, (cat)grp_size=1 /dev/dsk/dks1d4s2 (1100800 blks) show stat [subvolume] Displays statistics gathered for either all subvolumes or the specified subvolume. For example: xlv_mgr> show stat root.data Subvolume 0 (root.data): read ops: 668, read blocks: 286015 write ops: 2025, write blocks: 305508 The above example shows the number of read/write operations and read/write blocks going to the data subvolume of the volume root. The following example shows the statistics for subvolume 6 which has a striped volume element. xlv_mgr> show stat 6 Subvolume 6 (rickey.data): read ops: 300, read blocks: 2457600 write ops: 300, write blocks: 2457600 stripe ops: 600, total units: 19200 largest single i/o: 32 stripe units, frequency: 600 aligned < stripe width; ends on stripe unit: 0 aligned > stripe width; ends on stripe unit: 600 aligned = stripe width; ends on stripe unit: 0 aligned > or < stripe width; doesn't end on stripe unit: 0 unaligned < stripe width; ends on stripe unit: 0 unaligned > stripe width; ends on stripe unit: 0 unaligned = stripe width; doesn't end on stripe unit: 0 unaligned > or < stripe width; doesn't end on stripe unit: 0 A stripe width is composed of the product of the stripe unit and the number of disks comprising the stripe. The stripe unit determines the number of blocks to write to one disk before writing to the next. The stripe statistics are: stripe ops The number of operations issued to the subvolume's striped volume elements. Depending on the volume geometry, and whether or not the subvolume is plexed, this may not match the number of read and write operations. total units The cumulative count of all stripe units involved in stripe operations to the subvolume. largest single i/o Keeps track of the largest I/O size issued to the subvolume measured in stripe units. frequency The number of times the largest I/O was issued. There are three parameters for evaluating how an I/O spans a striped volume: aligned/unaligned Indicates whether or not the start address of the I/O begins on a stripe unit boundary. If an I/O doesn't start on a boundary, the stripe unit is used inefficiently. < stripe width > stripe width = stripe width Specifies if the transfer is less than a stripe width's worth of data, more than a stripe width, or an even multiple of the stripe width size. Transfers of full stripe widths are optimal since all disks are accessed evenly. Transfers of more or less than a stripe width are less optimal due to uneven drive access patterns. doesn't end on a stripe unit ends on a stripe unit When an I/O ends on a stripe unit, the next sequential I/O will start aligned on a stripe unit boundary. If it doesn't end on a stripe unit, then the stripe is used inefficiently since a sequential access will require that particular unit to be accessed multiple times. attach ve source dest-plex attach ve source volume.{data|log|rt}.N The command appends standalone volume element object source to the end of destination plex. This enables you to grow a plex or volume by adding a volume element to the end of a plex. You can use this in conjunction with xfs_growfs(1M) to grow an XFS filesystem without unmounting it. Suppose that you have a volume element, spareve, that contains a single disk partition /dev/dsk/dks1d4s2. The following command appends it to plex 0 of the data subvolume of volume db1: xlv_mgr> attach ve spareve db1.data.0 xlv_mgr> show object db1 VOL db1 (complete) VE db1.data.0.0 [active] start=0, end=1100799, (cat)grp_size=1 /dev/dsk/dks1d4s0 (1100800 blks) VE db1.data.0.1 [active] start=1100800, end=2201599, (cat)grp_size=1 /dev/dsk/dks1d4s1 (1100800 blks) VE db1.data.0.2 [active] start=2201600, end=3302399, (cat)grp_size=1 /dev/dsk/dks1d4s2 (1100800 blks) attach plex source volume.{data|log|rt} Appends standalone plex object source to existing volume volume. This command creates duplicate copies of the data on the volume for greater reliability. This operation is sometimes called mirroring. After the plex has been added, xlv_mgr automatically initiates a plex revive operation; this copies the data from the original XLV plexes to the newly added plex so that the plex holds the same data as the original plexes in the volume. The following appends a plex named plex2 to the data subvolume of volume db1: xlv_mgr> attach plex plex2 db1.data Use the show object command to display volume db1 and see that the disk partitions that were part of plex2 are now a component of db2. plex2 no longer exists as a standalone plex since it was merged into volume db1. insert ve source vol.{data|log|rt}.N insert ve source plex.N Inserts standalone volume element object source into the destination plex object. This enables you to add a volume element into a gap in a plex. xlv_mgr requires that the destination be a fully qualified XLV pathname (for example, movies.data.0). The pathname must specify the relative position within the plex to insert the volume element. (The first volume element in a plex is at position 0.) The plex to be operated on can be a standalone plex or a part of a volume. If the plex is part of a volume, the volume, subvolume, and plex must be specified. In the example below it is volume test. The following example inserts a volume element ve5 into a gap in the volume test. There is a gap because the first volume element starts at block number 76200. First display the configuration of test and ve5 before inserting ve5 into test. xlv_mgr> show object test VOL test (has holes) VE test.data.0.0 [active] start=76200, end=152399, (cat)grp_size=1 /dev/dsk/dks0d2s1 (76200 blks) VE test.data.0.1 [active] start=152400, end=228599, (cat)grp_size=1 /dev/dsk/dks0d2s2 (76200 blks) xlv_mgr> show object ve5 VE ve5 [empty] start=0, end=76199, (cat)grp_size=1 /dev/dsk/dks0d2s5 (76200 blks) xlv_mgr> insert ve ve5 test.data.0 xlv_mgr> show object test VOL test (complete) VE test.data.0.0 [stale] start=0, end=76199, (cat)grp_size=1 /dev/dsk/dks0d2s5 (76200 blks) VE test.data.0.1 [active] start=76200, end=152399, (cat)grp_size=1 /dev/dsk/dks0d2s1 (76200 blks) VE test.data.0.2 [active] start=152400, end=228599, (cat)grp_size=1 /dev/dsk/dks0d2s2 (76200 blks) detach [-force] ve plex.N ve-object detach [-force] ve volume.{data|log|rt}.N ve-object Removes specified volume element from its parent object and save it as ve-object. This command separates a volume element from its parent plex. This volume element can later be added to some other XLV object. The plex from which the volume element is removed can be a standalone plex or part of a volume. The detached volume element becomes a standalone XLV volume element object. You must specify the fully qualified pathname of the volume element to be detached and the name to be given to the detached volume element. The -force option forces the detach when the the parent object is missing any pieces. The detach operation leaves the volume element intact. detach [-force] plex volname.{data|log|rt}.plexno plexobject This command removes the specified plex from its parent object and saves it as plexobject. The -force option forces the detach when the the parent object is missing any pieces. This new standalone plex can later be added back to a volume via the attach plex command. The following example shows how to detach the first plex from a volume: xlv_mgr> detach plex db1.data.0 savedplex delete object name Deletes the object name. This command enables you to delete a volume, a standalone plex, or a standalone volume element. The XLV configuration is removed from the disks that make up the XLV object. Because the XLV configuration information is stored in the volume header (see vh(7M)), this command does not affect any user data that may have been written to the user disk partitions. delete all[_labels] An expert command, this command deletes the XLV labels from all disks on the local system. You might want to do this to initialize all the disks on a new system and ensure that there is no leftover XLV configuration information on the disks. Note that this is a very dangerous operation. Deleting the disk labels destroys all of the XLV objects on the system. delete label device_volume_header An expert command, this command deletes the XLV disk label from the named device_volume_header. change name object_name newname This command renames the given XLV object object_name to newname. The following example shows how to rename an object small to big. xlv_mgr> change name small big change nodename name object ... This command changes the nodename associated with the named objects. The following example shows how to set the node name for the volume db1 to homestead. xlv_mgr> change nodename homestead db1 change online vol.{data|log|rt}.N.N change offline vol.{data|log|rt}.N.N This command transitions the specified volume element online or offline. change stat on change stat off Expert commands to enable or disable statistics gathering. The default setting for statistics gathering is specified in the configuration file /var/sysgen/master.d/xlv. change type ve|plex|vol object This command changes the type of the given object. The following example shows how to change a standalone plex object sa_plex to a volume and then rename the volume to backup_vol. xlv_mgr> change type vol sa_plex xlv_mgr> change name sa_plex backup_vol change ve_start start_block sa_ve This command changes the start address of a standalone volume element. This is useful to aid in detaching a volume element from one object and arbitrarily inserting it in another. change plexmem max-pool-i/os growth% maximum-miss% [maximum-size] change vemem max-pool-i/os growth% maximum-miss% [maximum-size] An expert command, these two commands allow you to modify the parameters for the two XLV memory/buffer pools. When servicing an I/O request, there are two types of memory allocated. The first is for plexed volumes, if plexing support is present, and the second is for the underlying volume elements. If no memory available in the respective memory pool, XLV gets memory from the system general purpose memory allocator. By changing the maximum number of concurrent I/O max-pool-i/os, you can increase the efficiency of an I/O request because memory taken from the pool is more efficient than using the general purpose allocator. Changing the growth percentage growth% alters how fast the pool grows when the number of misses exceeds the miss percentage maximum-miss%. The maximum-size parameter is expressed in kilobyte units and is used to limit the size of the memory pool. Note that any changes to the pool parameters are only for life of the running system. The changes are not kept across system restarts. reset Reinitializes xlv_mgr data structures by rereading all the XLV configuration information from all the disks. reset stat [subvolume] An expert command to clear the statistics of either all subvolumes or the named subvolume. script [-write filename] object name script [-write filename] all Generates the required xlv_make commands to create the named object or all objects. When the -write option is specified, the xlv_make commands are saved into filename. help Displays a summary of xlv_mgr commands. ? Same as help. sh Fork a shell. quit Terminate this session. EXAMPLES These examples show the making of some XLV objects and the growing and plexing of a volume. The resulting volume configuration is saved to a xlv_make(1M) input file. Use xlv_make(1M) to create a volume named bigvolume. xlv_make> vol bigvolume bigvolume xlv_make> data bigvolume.data xlv_make> plex bigvolume.data.0 xlv_make> ve dks131d3s7 bigvolume.data.0.0 xlv_make> end Object specification completed xlv_make> exit Newly created objects will be written to disk. Is this what you want?(yes) yes Invoking xlv_assemble Create a standalone volume element spare_ve. # xlv_make xlv_make> ve spare_ve dks131d4s7 spare_ve xlv_make> end Object specification completed xlv_make> exit Newly created objects will be written to disk. Is this what you want?(yes) yes Invoking xlv_assemble Use the xlv_mgr attach ve command to grow the volume bigvolume by appending spare_ve to plex 0 of the data subvolume. # xlv_mgr xlv_mgr> attach ve spare_ve bigvolume.data.0 xlv_mgr> exit Create a standalone plex spare_plex which can be used to completely mirror the volume bigvolume. xlv_make> plex spare_plex spare_plex xlv_make> ve dks133d3s7 spare_plex.0 xlv_make> ve dks133d4s7 spare_plex.1 xlv_make> end Object specification completed xlv_make> exit Newly created objects will be written to disk. Is this what you want?(yes) yes Invoking xlv_assemble Use the xlv_mgr attach plex command to mirror the volume bigvolume by adding plex spare_plex to the data subvolume. # xlv_mgr xlv_mgr> attach plex spare_plex bigvolume.data xlv_mgr> exit Save the volume configuration to a file. # xlv_mgr xlv_mgr> script -write /etc/xlv_make.bigvolume.input object bigvolume xlv_mgr> exit SEE ALSO xlv_assemble(1M), xlv_make(1M), xlv_plexd(1M), xlv_shutdown(1M), xlv(7M). Tcl and the Tk Toolkit by John K. Ousterhout, Addison-Wesley, 1994. NOTES xlv_mgr operations modify both the XLV disk labels and the kernel data structures as appropriate. This means that you do not need to run xlv_assemble(1M) for your changes to take effect. The only exceptions are the XLV label deleting commands delete all_labels and delete label, which effect only the disk labels. xlv_mgr automatically initiates plex revive operations (see xlv_plexd(1M)) as required after adding a new plex to a volume or a volume element to a plexed volume. You should be root when running xlv_mgr. Otherwise you have limited functionality. Page 11