logo

Manual Pages


Table of Contents

NAME

disktest - Disk Test Environment

SYNOPSIS

disktest [ -B ] [ -t minutes ] [ -v ] [ adapter ]

disktest -T [ -t minutes ] [ -v ] adapter

disktest [ -R ] [ -W ] [ -A ] [ -WV ] [ -V ] [ -B ] [ -t minutes ] [ -n sects ] [ -v ] [ -s <shelf-list> ] [ -d <disk-list> ] [ -a <adapter-list> ]

DESCRIPTION

Use the disktest command to test all types of disks on an appliance. This command provides a report of the integrity of your storage environment. It is only available in maintenance mode. By default, it takes about 5 minutes to complete.

The -R option executes a sequential read test with optionally specified large block size (default is 1024kb per I/O).

The -W option executes a sequential write test with optionally specified large block size (default is 1024kb per I/O).

The -A option executes a test that alternates between writes and reads with optionally specified large block size (default is 1024kb per I/O). No data verification is peformed.

The -WV option executes a sequential write verify test which uses 4kb per I/O operation. This is identical to the way disktest would function with -V option on previous releases.

The -V option executes a sequential SCSI verify test which uses 10MB per operation. This test will run for one complete pass to verify all sectors on the disk regardless of -T option.

The -T option executes a test that alternates between writes and reads with varying I/O sizes. It also steps through permutations of shelves on the specified loop. If -t minutes is specified, each iteration of the test will run for the specified time. This test is a continuous test and will run until stopped via ^C.

The -n option is used to optionally specify the number of sectors to be read for each I/O of the -R,-A or -W option. The number of sectors used by the the -WV command is fixed at 8 (4kb) and cannot be altered. The number of sectors used by the -V command is fixed at 20480 (10MB) to increase throughput and cannot be altered.

The -d option allows for running disktest over a specific set of disks in the system by specifying a disk list of the form: <disk-name1> <disk-name2>

The -s option allows for running disktest over all disks contained in a specific shelf by specifying a shelf list of the form: <a>:<m> [<b>:<n> ...] where <m> and <n> are integer shelf ids and <a> and <b> are the PCI slot numbers of the Adapter(s) the shelves are connected to. (on board adapter is slot 0a) Hint: use fcadmin device_map to get slot locations.

The -a option allows for running disktest over a specific set of adapters in the system by specifying an adapter list of the form: <slot1> <slot2> ... <slotN>.

If the -v option is specified, the output is verbose.

If the -B option is specified, disks attached to a Fibre Channel loop via their B ports will also be tested.

By default, the test runs for about 5 minutes. However, if the [ -t minutes ] option is used, the test will run for the specified duration. If [ -t 0 ] is specified, the test will run CONTINUOUSLY until stopped with a ^C.

If the adapter or disk-list, adapter-list and shelf-list arguments are missing, all adapters and disks in the system are tested. Otherwise, only the specified adapter and disks attached to it are tested.

When finished, disktest prints out a report of the following values for each Fibre Channel adapter tested:

1. Number of times loss of synchronization was detected in that adapter's Fibre Channel loop.

2. Number of CRC errors found in Fibre Channel packets.

3. The total number of inbound and outbound frames seen by the adapter.

4. A "confidence factor" on a scale from 0 to 1 that indicates the health of your disk system as computed by the test. A value of 1 indicates that no errors were found. Any value less than 1 indicates there are problems in the Fibre Channel loop that are likely to intefere with the normal operation of your appliance. For more information see the Easy Installation Instructions for your specific filer or your storage shelf guide.

If the confidence factor is reported as less than 1, please go through the troubleshooting checklist for Fibre Channel loop problems in the document "Easy Installation Instructions for NetApp Filers" and re-run the disktest command after making any suggested modifications to your Fibre Channel setup. If the problem persists, please call your Customer Support telephone number.

The actual arithmetic that is used to compute the confidence factor is as follows:

The number of errors is obtained by adding the number of underrun, CRC, Synchronization and link failure errors with all errors weighted the same.

The allowable number of errors by the Fibrechannel protocol is calculated by adding fibre channel frames (inbound + outbound) and then multiplying by 2048 bytes per frame and dividing by the BER of 1e-12 converted to bytes at 1e-11.

The confidence factor is calculated as follows:

if total errors = 0 then confidence factor = 1.0

if total errors < allowable errors then confidence factor = 0.99

if total errors > allowable errors then confidence factor is decremented by .01 for each error seen which the protocol error rate does not allow.

When finished, disktest prints out a report of the following values for each adapter tested:

1. Number of Write operations performed on an adapter.

2. Number of Read operations performed on an adapter.

3. IOPS (I/O's per second) performed on an adapter.

4. Data rate in MB/S of the adapter.

5. Data transfer size per I/O operation on the adapter.

6. Number of soft (recovered) errors on the adapter.

7. Number of hard (unrecoverable) errors on the adapter.

8. A "confidence factor" on a scale from 0 to 1 that indicates the health of your disk system as computed by the test. A value of 1 indicates that no errors were found. Any value less than 1 indicates there are problems in the loop or bus or disk that are likely to intefere with the normal operation of your appliance. For more information see the Easy Installation Instructions for your specific filer or your storage shelf guide.

If the confidence factor is reported as less than 1, and a disk is reporting hard errors, you may want to proactively fail that disk or call your Customer Support telephone number.

The actual arithmetic that is used to compute the confidence factor is as follows:

The number of errors is obtained by adding the number of hard and soft errors from the disk with all errors weighted the same.

The allowable number of errors is zero for SCSI devices.

The confidence factor is calculated as follows:

if total errors = 0 then confidence factor = 1.0

if total errors > 0 then confidence factor is decremented by .01 for each error seen.

CLUSTER CONSIDERATIONS

In a clustered configuration, only disks on a filer's FCAL primary loop (the A loop) are tested, unless the -B option is specified. If -B is specified, disks on the B loop are tested as well.

EXAMPLES

The following command runs disktest for 5 minutes doing a sequential alternating write and read test in verbose mode on all adapters in the system, while testing only those disks which are attached via their A ports:

disktest -v

The following command runs disktest for an hour doing a sequential write test in verbose mode, using 1024kb I/O blocks while testing disks attached to adapter 8 via both A and B ports:

disktest -W -v -B -t 60 -a 8

The following command runs disktest for 5 minutes doing a sequential read test on all disks in shelf 0 on adapter 7.

disktest -R -s 7:0

The following command runs disktest continuously (until stopped) doing a sequential write test of 512kb I/O's to all disks on shelf 1 on adapter 7, shelf 2 on adapter 7, disks 7.0 and 7.1 and all disks on adapter 8.

disktest -W -n 1024 -t 0 -d 7.0 7.1 -s 7:1 7:2 -a 8

The following command runs disktest continuously (until stopped) doing an alternating sequential write/read test with varying I/O sizes across all shelf permutations in the loop attached to adapter 7 for 4 minutes on each iteration.

disktest -T -t 4 7


Table of Contents