vxstat(1M)

vxstat - Volume Manager statistics management utility

Synopsis

vxstat [ -g diskgroup ] [ -i interval [ -c count ] ] [ -f fields ] [ -r ] [ -psvd ] [ object ... ]

Description

The vxstat utility prints and resets statistics information on one or more volumes, plexes, subdisks, or disks.

The vxstat utility reads statistics from the volume device files in the directory /dev/vx/rdsk and prints them to standard output. These statistics represent volume, plex, subdisk, and disk activity since boot time. If no object operands are given, then statistics from all volumes in the configuration database are reported. object can be the name of a volume, plex, subdisk, or disk.

Options

-g diskgroup

Select records from the specified disk group. The diskgroup option argument can be either a disk group name or disk group ID.

-i interval

Print the change in volume statistics that occurs after every interval seconds. The first interval is assumed to encompass the entire previous history of objects. Subsequent displays will show statistics with a zero value if there has been no change since the previous interval.

-c count

Stop after printing interval statistics count times.

-f fields

Select the display of statistics collected. The following options are available:

s

Statistics on read and write operations. Displays six fields: the number of read operations, the number of write operations, the number of blocks read, the number of blocks written, the average time spent on read operations in the interval, and the average time spent on write operations in the interval. These statistics are displayed as the default output format.

a

Statistics on atomic copies performed (has meaning only for mirrored volumes). Displays the number of operations, number of blocks, and the average time spent per operation.

v

Statistics on verified reads and writes (has meaning only for mirrored volumes). Displays six fields: the number of verified read operations, the number of verified write operations, the number of blocks read, the number of blocks written, the average time spent on verified read operations in the interval, and the average time spent on verified write operations in the interval.

c

Statistics on corrected (fixed) read operations (has meaning only for mirrored or RAID-5 volumes). Displays the number of fixed read and write operations.

NOTE: Currently, only read operations are ever corrected, so the number of fixed writes will always be zero.

f

Displays the number of failed read and write operations.

b

Displays the statistics on read-writeback mirror consistency recovery operations (has meaning only for mirrored volumes). Displays three fields: the number of read- writeback operations, the number of blocks involved in read-writeback operations, and the average time for completing a read-writeback operation. While in recovery mode, most read operations to a mirrored volume invoke read- writeback consistency recovery.

F

Statistics for full-stripe writes on a RAID-5 volume. The number of operations represents the number of write operations within a stripe that were conducted as a full-stripe write optimization. Full-stripe writes represent considerably less overhead than read-modify-writes in terms of overall I/O time, latency and CPU overhead. The total number of blocks represents the total size of the written data and the average time is the time taken for a full-stripe write operation. Since the I/O may be larger then a single stripe, more than one stripe operation may be seen for a single logical I/O request.

M

Read-modify-write statistics. Each operation represents a read-modify write operation performed within a stripe. I/O crossing a stripe boundary will be represented by more than a single read-modify write operation. The number of blocks counted represents only the size of the requested write. The read portion of the I/O can be derived.

W

Reconstruct write statistics. Each operation counted is for a reconstruct write operation performed as an optimization of a write operation within a stripe. The number of blocks counted represents the count of data blocks written not including parity or read operations.

R

Reconstruct read operations. Each operation is a separate reconstruct read operation. A single stripe read or write operation can lead to numerous reconstruct read operations since each reconstruction takes place at the subdisk level. A detached column can consist of several subdisks each of which will lead to a reconstruct read operation.

0

Statistics for the VOL_R5_ZERO operation. Each operation represents one call to the VOL_R5_ZERO ioctl. The number of blocks is based on the number of zero'd blocks written to the array. The average time is the time taken to complete the entire ioctl operation.

S

Statistics for the VOL_R5_RESYNC operation. Each operation count represents one call to the VOL_R5_RESYNC ioctl. The number of blocks represents the resulting number of blocks that were written to the parity regions as part of the resynchronization of parity. The read operations are not counted towards the total.

C

Statistics for the VOL_R5_RECOVER operation. Each operation count represents one call to the VOL_R5_RECOVER ioctl. The number of blocks represents the resulting number of blocks that were written to the missing column region as part of the data recovery. The read operations are not counted towards the total.

V

Statistics for the VOL_R5_VERIFY operation. Each operation count represents one call to the VOL_R5_VERIFY ioctl. The number of blocks represents the resulting number of blocks that were read as part of the RAID-5 stripes consistency verification.

-r

Reset statistics instead of printing them. This option will follow the same selection rules as printing for any type selection arguments or for any named objects. If an interval was specified on the command line, then the first set of statistics will not be printed since they will have been reset to zero. Subsequent activity will cause printing of statistics as normal.

-v

Display statistics for volumes on the objects specified on the command line. For an object that is a plex or a subdisk, displays information about the volume with which the object is associated. If an object supplied is a disk, then any volumes that occupy any part of the disk will be selected.

-p

Display statistics for plexes on the object specified on the command line. For subdisk objects, displays information about a plex with which it is associated.

-s

Display statistics for subdisks on the objects specified on the command line.

-d

Display statistics for disks on which the object specified on the command line is fully or partially located.

Output format

Summary statistics for each object are printed in one-line output records, preceded by two header lines. The output line consists of blank-separated fields for the object type, object name (standard), and the fields requested by the -f switch in the order they are specified on the command line.

If the -i interval option was supplied, then statistics will be prefaced with a time-stamp showing the current local time on the system.

Exit codes

The vxstat utility exits with a nonzero status if the attempted operation fails. A nonzero exit code is not a complete indicator of the problems encountered, but rather denotes the first condition that prevented further execution of the utility. See vxintro(1M) for a list of standard exit codes.

Examples

To display statistics for all subdisks associated with all volumes, use the command:

	vxstat -s

To display statistics for the plexes and subdisks of a volume named blop, use the following:

	vxstat -ps blop

To reset all statistics for a disk group named foodg, type the following command:

	vxstat -g foodg -r

To display 5 sets of disk statistics at 10 second intervals, use the following:

	vxstat -i 10 -c 5 -d

References

vxintro(1M), vxtrace(1M), vxtrace(7)