Disk performance and health

About monitoring disk performance and health

Data ONTAP continually monitors disks to assess their performance and health. When Data ONTAP encounters specific activities on a disk, it will take corrective action by either taking a disk offline temporarily or by taking it out of service to run further tests. When this occurs, the disk is in the maintenance center.

When Data ONTAP takes disks offline temporarily

Data ONTAP temporarily stops I/O activity to a disk and takes a disk offline when

You update disk firmware
ATA disks take a long time to recover from a bad media patch

While the disk is offline, Data ONTAP reads from other disks within the RAID group while writes are logged. The offline disk is brought back online after re-synchronization is complete. This process generally takes a few minutes and incurs a negligible performance impact. For ATA disks, this reduces the probability of forced disk failures due to bad media patches or transient errors because taking a disk offline provides a software-based mechanism for isolating faults in drives and for performing out-of-band error recovery.

The disk offline feature is only supported for spares and data disks within RAID-DP and mirrored-RAID4 aggregates. A disk can be taken offline only if its containing RAID group is in a normal state and the plex or aggregate is not offline.

You view the status of disks with the aggr status -r or aggr status -s commands, as shown in the following examples. You can see what disks are offline with either option.

Note
For backward compatibility, you can also use the vol status -r or vol status -s commands.

Example 1

system> aggr status -r aggrA

Aggregate aggrA (online, raid4-dp degraded) (block checksums)

Plex /aggrA/plex0 (online, normal, active)

RAID group /aggrA/plex0/rg0 (degraded)

RAID Disk Device HA SHELF BAY CHAN Pool Type RPM Used (MB/blks) 
Phys (MB/blks)

--------- ------ ------------- ---- ---- ---- ----- -------------- 
--------------

parity 8a.20 8a 1 4 FC:A - FCAL 10000 1024/2097152 1191/2439568 

data 6a.36 6a 2 4 FC:A - FCAL 10000 1024/2097152 1191/2439568 

data 6a.19 6a 1 3 FC:A - FCAL 10000 1024/2097152 1191/2439568 

data 8a.23 8a 1 7 FC:A - FCAL 10000 1024/2097152 1191/2439568 
(offline)

Example 2

system> aggr status -s

Spare disks

RAID Disk Device HA SHELF BAY CHAN Pool Type RPM Used (MB/blks) 
Phys (MB/blks)

--------- ------ ------------- ---- ---- ---- ----- -------------- 
--------------

Spare disks for block or zoned checksum traditional volumes or 
aggregates

spare 8a.24 8a 1 8 FC:A - FCAL 10000 1024/2097152 1191/2439568 

spare 8a.25 8a 1 9 FC:A - FCAL 10000 1024/2097152 1191/2439568 

spare 8a.26 8a 1 10 FC:A - FCAL 10000 1024/2097152 1191/2439568 
(offline)

spare 8a.27 8a 1 11 FC:A - FCAL 10000 1024/2097152 1191/2439568 

spare 8a.28 8a 1 12 FC:A - FCAL 10000 1024/2097152 1191/2439568

When Data ONTAP takes a disk out of service

When Data ONTAP detects disk errors, it takes corrective action. For example, if a disk experiences a number of errors that exceed predefined thresholds for that disk type, Data ONTAP takes one of the following actions:

If the disk.maint_center.spares_check option is set to on (which it is by default) and there are two or more spares available, Data ONTAP takes the disk out of service and assigns it to the maintenance center for data management operations and further testing,
If the disk.maint_center.spares_check option is set to on and there are less than two spares available, Data ONTAP does not assign the disk to the maintenance center. It simply fails the disk.
If the disk.maint_center.spares_check option is set to off, Data ONTAP assigns the disk to the maintenance center without checking the number of available spares.
Note
The disk.main_center.spares_check option has no affect on putting disks into the maintenance center from the command line interface.

Once the disk is in the maintenance center, it is subjected to a number of tests, depending on what type of disk it is. If the disk passes all of the tests, it is returned to the spare pool. If it is ever sent back to the maintenance center, it is automatically failed. If a disk doesn't pass the tests the first time, it is automatically failed.

Data ONTAP also informs you of these activities by sending messages to

The console
A log file at /etc/maintenance.log
A binary file that is sent with weekly AutoSupport messages

This feature is controlled by the disk.maint_center.enable option. It is on by default.

Manually running maintenance tests

You can initiate maintenance tests on a disk by using the disk maint start command. The following table summarizes how to use this command.

disk maint parameter	Information displayed
disk maint list	Shows all of the available tests.
disk maint start [-t test_list] [-c cycle_count] [-f] [-i] -d disk_list	Starts the test. `-t` test_list specifies which tests to run. The default is all. `-c` count_cycle specifies the number of cycles the tests will run on the disk. The default is 1. `-f` suppresses the warning message and forces execution of the command without confirmation (data disks only). If this option is not specified, the command issues a warning message and waits for confirmation before proceeding. `-i` instructs Data ONTAP to immediately remove the disk (data disks only) from the RAID group and begin the maintenance tests. As as a result, the RAID group enters degraded mode. If a suitable spare disk is available, the contents of the removed disk will be reconstructed onto that spare disk. If this option is not specified, Data ONTAP marks the disk as pending. If an appropriate spare is available, it is selected for Rapid RAID Recovery, and the data disk is copied to the spare. After the copy is completed, the data disk is removed form the RAID configuration and the testing begins. `-d disk_list` specifies a list of disks to run the tests on.
disk maint status [-v] [disk_list]	Shows the status of the disks in the maintenance center. `-v` specifies verbose. disk_list specifies a list of disks in the maintenance center to display the status of. The default is all.
`disk maint abort` disk_list	Stops the tests that are running on a disk in the maintenance center. If the specified disks were ones that you initiated the test on, they are returned to the spare pool. Those that were sent to the maintenance center by Data ONTAP are failed if aborted before testing is completed.