NAME

metatab − SPARCserver Manager metadisk configuration database

SYNOPSIS

/etc/metatab

AVAILABILITY

This configuration file is used by the Sun SPARCserver Manager (SSM) software only. 

DESCRIPTION

Note: The metatab(5) man page is included in this manual in the event your site is using the SPARCserver Manager (SSM) software package.   This man page is included here for historical purposes only.  The file is not active under DiskSuite.

/etc/metatab is the database used by the metadisk.  Each metadisk metapartition is a minor device of the metadisk device, and is represented by one line in /etc/metatab.  Each metapartition is either a concatenation of up to four real disk partitions, or a metamirror comprised of one or two regular metapartitions. 

Metapartitions specified in /etc/metatab are activated or set up by the /usr/etc/metainit(8) command. 

Each line in /etc/metatab is of the form: special rows 1 real1 1 real2 ... 1 realn which defines a disk concatenation of real partitions from real1 to realn, with its size set to be the sum of the sizes of all the real partitions, or special -m special options which defines a metamirror that consists of a single mirrored metapartition, special, with its size set to be equal to the size of the mirrored metapartition, or special -m special1 special2 options which defines a metamirror that consists of two mirrored metapartitions, special1 and special2, with its size set to be the smaller of the two mirrored metapartitions. 

Each of the mirrored metapartitions have to be defined separately in /etc/metatab like other metapartitions. 

The fields have the following meaning:

special
is the name of a metapartition, for example /dev/md0a, /dev/md2b . . ., etc. A total of 24 metapartitions can be configured and used. They are /dev/md[0-2][a-h]. 

rows is the number of rows in the metapartition.  Each metapartition may consist of a certain number of rows of real partitions.  The maximum allowable number of rows is four. 

real is the name of the real disk partition.  /usr/etc/metainit will complain if a real disk partition appears in more than one row of a metapartition, or in more than one metapartition. It will also complain if rows are not followed by a sufficient number of real partition names. 

The pattern 1 real is repeated up to four times, once for each row in the metapartition. 
 

options:

The following flags control the mode of operation for disk mirroring:

-s means dispatch writes to the second mirrored metapartition (if it exists) after writes to the first mirrored metapartition has completed. Without this flag, all writes to the second mirrored metapartition will be dispatched in parallel with corresponding writes to the first mirrored metapartition. 

-g means that when doing a read, if the disk block to be read is on the first half of the metamirror, read it from the first mirrored metapartition. Otherwise read it from the second mirrored metapartition. 

-r means read only from the first mirrored metapartition. This flag should not be used with the -g flag. When neither flag is specified, the metamirror will alternate reads between the first and the second metapartitions. 

-e number
specifies an error threshold for the number of read errors that will be tolerated before taking a faulty mirrored metapartition off-line. Note that a mirrored metapartition will be taken offline as soon as the first write error occurs.

single_digit
May be specified as the last entry of the option fields. It designates the pass number used for mirror resync during a system reboot. Mirror resync for mirrored pairs with smaller pass numbers will be scheduled before those with larger pass numbers. Mirror resync for metamirrors with equal pass number will be run concurrently. A 0 (zero) means to skip the mirror-resync during reboot.  If the pass number is not specified, it defaults to 1. 

For further details, see examples below. 

System V File-Creation Semantics

None

EXAMPLES

All IPI drives in the following examples have the same size of 2006130 sectors. 

Concatenation:
This example shows a metapartition, /dev/md0c, consisting of a concatenation of two disks. /dev/md0c 2 1 /dev/id000c 1 /dev/id001c Since the first disk sector in /dev/id001c contains a disk label, it will be skipped by the metadisk.  So /dev/md0c will contain 2006130∗2-1 blocks, which is the total number of blocks on the two IPI disks, minus one. The first disk block from /dev/id001c is not used by /dev/md0c. The second disk block on /dev/id001c corresponds to block 2006130 on /dev/md0c.

Mirroring:
This example shows a configuration of a metamirror, /dev/md2a, comprised of two metapartitions, /dev/md0a and /dev/md0b.

/dev/md2a -m /dev/md0a /dev/md0b -e 10 2
/dev/md0a 1 1 /dev/id001c
/dev/md0b 1 1 /dev/id010c

Any write request to metapartition /dev/md2a will be written to both /dev/md0a and /dev/md0b; data will be duplicated on /dev/id001c and /dev/id010c.
 
Even numbered blocks will be read from /dev/id001c. Odd numbered blocks will be read from /dev/id010c. Therefore, the IO load for the reads will be evenly distributed between the two disks.
 
You should be aware of some performance implications when using disk mirroring: to minimize the performance penalty of disk mirroring, you should put all mirrored metapartitions on different controllers and you should not use the -s flag.  Then the writes will be dispatched to the mirrored metapartitions in parallel, and the elapsed time for completing a write will be reduced.

Concatenation+Mirroring:
In this example, two metapartitions are defined, each consisting of a concatenation of two IPI disks. Then a metamirror is defined on top of the two metapartitions.

/dev/md1c 2 1 /dev/id002c 1 /dev/id003c
/dev/md1d 2 1 /dev/id010c 1 /dev/id011c
/dev/md2c -m /dev/md1c /dev/md1d -g -s

The size of /dev/md2c is 4012259 blocks, the same as /dev/md1c and /dev/md1d.
 
The -s flag means that writes to /dev/md1d will only be done after writes to /dev/md1c are completed.  This means that /dev/md1c may be more up-to-date than /dev/md1d at any given time. 
 
The -g flag means that blocks are read only from /dev/id002c and /dev/id011c. 
 
Logic indicates that if one of the real partitions has a disk failure, you should replace the failing disk as soon as possible, and run the metadisk utility /usr/etc/metasync, which allows resyncing of the mirrored pair with minimal impact on the availability of the metapartition. 

FILES

/etc/metatab
database used by the Sun SPARCserver Manager

SEE ALSO

fstab(5) md.conf(5)

Sun Online: DiskSuite Administration Guide
SPARCserver Manager User Guide

BUGS

Recursive mirroring is not allowed; that is, a metamirror cannot be a component in the definition of another metamirror. 

Sun Release 4.1  —  Last change: 23 July 1991