OpenBSD manual page server

Configuration file

The format of the configuration file is complex, and only an abbreviated treatment is given here. In the configuration files, a ‘#’ indicates the beginning of a comment.

There are 4 required sections of a configuration file, and 2 optional sections. Each section begins with a ‘START’, followed by the section name, and the configuration parameters associated with that section. The first section is the ‘array’ section, and it specifies the number of rows, columns, and spare disks in the RAID set. For example:

START array
1 3 0

indicates an array with 1 row, 3 columns, and 0 spare disks. Note that although multi-dimensional arrays may be specified, they are NOT supported in the driver.

The second section, the ‘disks’ section, specifies the actual components of the device. For example:

START disks
/dev/sd0e
/dev/sd1e
/dev/sd2e

specifies the three component disks to be used in the RAID device. If any of the specified drives cannot be found when the RAID device is configured, then they will be marked as ‘failed’, and the system will operate in degraded mode. Note that it is imperative that the order of the components in the configuration file does not change between configurations of a RAID device. Changing the order of the components will result in data loss if the set is configured with the -C option. In normal circumstances, the RAID set will not configure if only -c is specified, and the components are out-of-order.

The next section, which is the ‘spare’ section, is optional, and, if present, specifies the devices to be used as ‘hot spares’ -- devices which are on-line, but are not actively used by the RAID driver unless one of the main components fail. A simple ‘spare’ section might be:

START spare
/dev/sd3e

for a configuration with a single spare component. If no spare drives are to be used in the configuration, then the ‘spare’ section may be omitted.

The next section is the ‘layout’ section. This section describes the general layout parameters for the RAID device, and provides such information as sectors per stripe unit, stripe units per parity unit, stripe units per reconstruction unit, and the parity configuration to use. This section might look like:

START layout
# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level
32 1 1 5

The sectors per stripe unit specifies, in blocks, the interleave factor; i.e. the number of contiguous sectors to be written to each component for a single stripe. Appropriate selection of this value (32 in this example) is the subject of much research in RAID architectures. The stripe units per parity unit and stripe units per reconstruction unit are normally each set to 1. While certain values above 1 are permitted, a discussion of valid values and the consequences of using anything other than 1 are outside the scope of this document. The last value in this section (5 in this example) indicates the parity configuration desired. Valid entries include:

0

RAID level 0. No parity, only simple striping.

1

RAID level 1. Mirroring. The parity is the mirror.

4

RAID level 4. Striping across components, with parity stored on the last component.

5

RAID level 5. Striping across components, parity distributed across all components.

There are other valid entries here, including those for Even-Odd parity, RAID level 5 with rotated sparing, Chained declustering, and Interleaved declustering, but as of this writing the code for those parity operations has not been tested with OpenBSD.

The next required section is the ‘queue’ section. This is most often specified as:

START queue
fifo 100

where the queuing method is specified as FIFO (First-In, First-Out), and the size of the per-component queue is limited to 100 requests. Other queuing methods may also be specified, but a discussion of them is beyond the scope of this document.

The final section, the ‘debug’ section, is optional. For more details on this the reader is referred to the RAIDframe documentation discussed in the HISTORY section. See EXAMPLES for a more complete configuration file example.

Initialization and Configuration

The initial step in configuring a RAID set is to identify the components that will be used in the RAID set. All components should be the same size. Each component should have a disklabel type of FS_RAID, and a typical disklabel entry for a RAID component might look like:

f:  1800000  200495     RAID              # (Cyl.  405*- 4041*)

While FS_BSDFFS (e.g. 4.2BSD) will also work as the component type, the type FS_RAID (e.g. RAID) is preferred for RAIDframe use, as it is required for features such as auto-configuration. As part of the initial configuration of each RAID set, each component will be given a ‘component label’. A ‘component label’ contains important information about the component, including a user-specified serial number, the row and column of that component in the RAID set, the redundancy level of the RAID set, a 'modification counter', and whether the parity information (if any) on that component is known to be correct. Component labels are an integral part of the RAID set, since they are used to ensure that components are configured in the correct order, and used to keep track of other vital information about the RAID set. Component labels are also required for the auto-detection and auto-configuration of RAID sets at boot time. For a component label to be considered valid, that particular component label must be in agreement with the other component labels in the set. For example, the serial number, ‘modification counter’, number of rows and number of columns must all be in agreement. If any of these are different, then the component is not considered to be part of the set. See raid(4) for more information about component labels.

Once the components have been identified, and the disks have appropriate labels, raidctl is then used to configure the raid(4) device. To configure the device, a configuration file which looks something like:

START array
# numRow numCol numSpare
1 3 1

START disks
/dev/sd1e
/dev/sd2e
/dev/sd3e

START spare
/dev/sd4e

START layout
# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_5
32 1 1 5

START queue
fifo 100

is created in a file. The above configuration file specifies a RAID 5 set consisting of the components /dev/sd1e, /dev/sd2e, and /dev/sd3e, with /dev/sd4e available as a ‘hot spare’ in case one of the three main drives should fail. A RAID 0 set would be specified in a similar way:

START array
# numRow numCol numSpare
1 4 0

START disks
/dev/sd10e
/dev/sd11e
/dev/sd12e
/dev/sd13e

START layout
# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_0
64 1 1 0

START queue
fifo 100

In this case, devices /dev/sd10e, /dev/sd11e, /dev/sd12e, and /dev/sd13e are the components that make up this RAID set. Note that there are no hot spares for a RAID 0 set, since there is no way to recover data if any of the components fail.

For a RAID 1 (mirror) set, the following configuration might be used:

START array
# numRow numCol numSpare
1 2 0

START disks
/dev/sd20e
/dev/sd21e

START layout
# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_1
128 1 1 1

START queue
fifo 100

In this case, /dev/sd20e and /dev/sd21e are the two components of the mirror set. While no hot spares have been specified in this configuration, they easily could be, just as they were specified in the RAID 5 case above. Note as well that RAID 1 sets are currently limited to only 2 components. At present, n-way mirroring is not possible.

The first time a RAID set is configured, the -C option must be used:

# raidctl -C raid0.conf raid0

where ‘raid0.conf’ is the name of the RAID configuration file. The -C forces the configuration to succeed, even if any of the component labels are incorrect. The -C option should not be used lightly in situations other than initial configurations, as if the system is refusing to configure a RAID set, there is probably a very good reason for it. After the initial configuration is done (and appropriate component labels are added with the -I option) then raid0 can be configured normally with:

# raidctl -c raid0.conf raid0

When the RAID set is configured for the first time, it is necessary to initialize the component labels, and to initialize the parity on the RAID set. Initializing the component labels is done with:

# raidctl -I 112341 raid0

where ‘112341’ is a user-specified serial number for the RAID set. This initialization step is required for all RAID sets. Also, using different serial numbers between RAID sets is strongly encouraged, as using the same serial number for all RAID sets will only serve to decrease the usefulness of the component label checking.

Initializing the RAID set is done via the -i option. This initialization MUST be done for all RAID sets, since among other things it verifies that the parity (if any) on the RAID set is correct. Since this initialization may be quite time-consuming, the -v option may be also used in conjunction with -i:

# raidctl -iv raid0

This will give more verbose output on the status of the initialization:

Initiating re-write of parity
Parity Re-write status:
 10% |****                                   | ETA:    06:03 /

The output provides a ‘Percent Complete’ in both a numeric and graphical format, as well as an estimated time to completion of the operation.

Since it is the parity that provides the ‘redundancy’ part of RAID, it is critical that the parity is correct as much as possible. If the parity is not correct, then there is no guarantee that data will not be lost if a component fails.

Once the parity is known to be correct, it is then safe to perform disklabel(8), newfs(8), or fsck(8) on the device or its filesystems, and then to mount the filesystems for use.

Under certain circumstances (e.g. the additional component has not arrived, or data is being migrated off of a disk destined to become a component) it may be desirable to configure a RAID 1 set with only a single component. This can be achieved by configuring the set with a physically existing component (as either the first or second component) and with a ‘fake’ component. In the following:

START array
# numRow numCol numSpare
1 2 0

START disks
/dev/sd6e
/dev/sd0e

START layout
# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_1
128 1 1 1

START queue
fifo 100

/dev/sd0e is the real component, and will be the second disk of a RAID 1 set. The component /dev/sd6e, which must exist, but have no physical device associated with it, is simply used as a placeholder. Configuration (using -C and -I 12345 as above) proceeds normally, but initialization of the RAID set will have to wait until all physical components are present. After configuration, this set can be used normally, but will be operating in degraded mode. Once a second physical component is obtained, it can be hot-added, the existing data mirrored, and normal operation resumed.

Maintenance of the RAID set

After the parity has been initialized for the first time, the command:

# raidctl -p raid0

can be used to check the current status of the parity. To check the parity and rebuild it necessary (for example, after an unclean shutdown) the command:

# raidctl -P raid0

is used. Note that re-writing the parity can be done while other operations on the RAID set are taking place (e.g. while doing an fsck(8) on a file system on the RAID set). However: for maximum effectiveness of the RAID set, the parity should be known to be correct before any data on the set is modified.

To see how the RAID set is doing, the following command can be used to show the RAID set's status:

# raidctl -s raid0

The output will look something like:

Components:
           /dev/sd1e: optimal
           /dev/sd2e: optimal
           /dev/sd3e: optimal
Spares:
           /dev/sd4e: spare
Parity status: clean
Reconstruction is 100% complete.
Parity Re-write is 100% complete.
Copyback is 100% complete.

This indicates that all is well with the RAID set. Of importance here are the component lines which read ‘optimal’, and the ‘Parity status’ line which indicates that the parity is up-to-date. Note that if there are file systems open on the RAID set, the individual components will not be ‘clean’ but the set as a whole can still be clean.

The -v option may be also used in conjunction with -s:

# raidctl -sv raid0

In this case, the components' label information (see the -g option) will be given as well:

Components:
           /dev/sd1e: optimal
           /dev/sd2e: optimal
           /dev/sd3e: optimal
Spares:
           /dev/sd4e: spare
Component label for /dev/sd1e:
   Row: 0 Column: 0 Num Rows: 1 Num Columns: 3
   Version: 2 Serial Number: 13432 Mod Counter: 65
   Clean: No Status: 0
   sectPerSU: 32 SUsPerPU: 1 SUsPerRU: 1
   RAID Level: 5  blocksize: 512 numBlocks: 1799936
   Autoconfig: No
   Last configured as: raid0
Component label for /dev/sd2e:
   Row: 0 Column: 1 Num Rows: 1 Num Columns: 3
   Version: 2 Serial Number: 13432 Mod Counter: 65
   Clean: No Status: 0
   sectPerSU: 32 SUsPerPU: 1 SUsPerRU: 1
   RAID Level: 5  blocksize: 512 numBlocks: 1799936
   Autoconfig: No
   Last configured as: raid0
Component label for /dev/sd3e:
   Row: 0 Column: 2 Num Rows: 1 Num Columns: 3
   Version: 2 Serial Number: 13432 Mod Counter: 65
   Clean: No Status: 0
   sectPerSU: 32 SUsPerPU: 1 SUsPerRU: 1
   RAID Level: 5  blocksize: 512 numBlocks: 1799936
   Autoconfig: No
   Last configured as: raid0
Parity status: clean
Reconstruction is 100% complete.
Parity Re-write is 100% complete.
Copyback is 100% complete.

To check the component label of /dev/sd1e, the following is used:

# raidctl -g /dev/sd1e raid0

The output of this command will look something like:

Component label for /dev/sd1e:
   Row: 0 Column: 0 Num Rows: 1 Num Columns: 3
   Version: 2 Serial Number: 13432 Mod Counter: 65
   Clean: No Status: 0
   sectPerSU: 32 SUsPerPU: 1 SUsPerRU: 1
   RAID Level: 5  blocksize: 512 numBlocks: 1799936
   Autoconfig: No
   Last configured as: raid0

Dealing with Component Failures

If for some reason (perhaps to test reconstruction) it is necessary to pretend a drive has failed, the following will perform that function:

# raidctl -f /dev/sd2e raid0

The system will then be performing all operations in degraded mode, where missing data is re-computed from existing data and the parity. In this case, obtaining the status of raid0 will return (in part):

Components:
           /dev/sd1e: optimal
           /dev/sd2e: failed
           /dev/sd3e: optimal
Spares:
           /dev/sd4e: spare

Note that with the use of -f a reconstruction has not been started. To both fail the disk and start a reconstruction, the -F option must be used:

# raidctl -F /dev/sd2e raid0

The -f option may be used first, and then the -F option used later, on the same disk, if desired. Immediately after the reconstruction is started, the status will report:

Components:
           /dev/sd1e: optimal
           /dev/sd2e: reconstructing
           /dev/sd3e: optimal
Spares:
           /dev/sd4e: used_spare
[...]
Parity status: clean
Reconstruction is 10% complete.
Parity Re-write is 100% complete.
Copyback is 100% complete.

This indicates that a reconstruction is in progress. To find out how the reconstruction is progressing the -S option may be used. This will indicate the progress in terms of the percentage of the reconstruction that is completed. When the reconstruction is finished the -s option will show:

Components:
           /dev/sd1e: optimal
           /dev/sd2e: spared
           /dev/sd3e: optimal
Spares:
           /dev/sd4e: used_spare
[...]
Parity status: clean
Reconstruction is 100% complete.
Parity Re-write is 100% complete.
Copyback is 100% complete.

At this point there are at least two options. First, if /dev/sd2e is known to be good (i.e. the failure was either caused by -f or -F, or the failed disk was replaced), then a copyback of the data can be initiated with the -B option. In this example, this would copy the entire contents of /dev/sd4e to /dev/sd2e. Once the copyback procedure is complete, the status of the device would be (in part):

Components:
           /dev/sd1e: optimal
           /dev/sd2e: optimal
           /dev/sd3e: optimal
Spares:
           /dev/sd4e: spare

and the system is back to normal operation.

The second option after the reconstruction is to simply use /dev/sd4e in place of /dev/sd2e in the configuration file. For example, the configuration file (in part) might now look like:

START array
1 3 0

START drives
/dev/sd1e
/dev/sd4e
/dev/sd3e

This can be done as /dev/sd4e is completely interchangeable with /dev/sd2e at this point. Note that extreme care must be taken when changing the order of the drives in a configuration. This is one of the few instances where the devices and/or their orderings can be changed without loss of data! In general, the ordering of components in a configuration file should never be changed.

If a component fails and there are no hot spares available on-line, the status of the RAID set might (in part) look like:

Components:
           /dev/sd1e: optimal
           /dev/sd2e: failed
           /dev/sd3e: optimal
No spares.

In this case there are a number of options. The first option is to add a hot spare using:

# raidctl -a /dev/sd4e raid0

After the hot add, the status would then be:

Components:
           /dev/sd1e: optimal
           /dev/sd2e: failed
           /dev/sd3e: optimal
Spares:
           /dev/sd4e: spare

Reconstruction could then take place using -F as describe above.

A second option is to rebuild directly onto /dev/sd2e. Once the disk containing /dev/sd2e has been replaced, one can simply use:

# raidctl -R /dev/sd2e raid0

to rebuild the /dev/sd2e component. As the rebuilding is in progress, the status will be:

Components:
           /dev/sd1e: optimal
           /dev/sd2e: reconstructing
           /dev/sd3e: optimal
No spares.

and when completed, will be:

Components:
           /dev/sd1e: optimal
           /dev/sd2e: optimal
           /dev/sd3e: optimal
No spares.

In circumstances where a particular component is completely unavailable after a reboot, a special component name will be used to indicate the missing component. For example:

Components:
           /dev/sd2e: optimal
          component1: failed
No spares.

indicates that the second component of this RAID set was not detected at all by the auto-configuration code. The name ‘component1’ can be used anywhere a normal component name would be used. For example, to add a hot spare to the above set, and rebuild to that hot spare, the following could be done:

# raidctl -a /dev/sd3e raid0
# raidctl -F component1 raid0

at which point the data missing from ‘component1’ would be reconstructed onto /dev/sd3e.

RAID on RAID

RAID sets can be layered to create more complex and much larger RAID sets. A RAID 0 set, for example, could be constructed from four RAID 5 sets. The following configuration file shows such a setup:

START array
# numRow numCol numSpare
1 4 0

START disks
/dev/raid1e
/dev/raid2e
/dev/raid3e
/dev/raid4e

START layout
# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_0
128 1 1 0

START queue
fifo 100

A similar configuration file might be used for a RAID 0 set constructed from components on RAID 1 sets. In such a configuration, the mirroring provides a high degree of redundancy, while the striping provides additional speed benefits.

Auto-configuration and Root on RAID

RAID sets can also be auto-configured at boot. To make a set auto-configurable, simply prepare the RAID set as above, and then do a:

to turn on auto-configuration for that set. To turn off auto-configuration, use:

RAID sets which are auto-configurable will be configured before the root file system is mounted. These RAID sets are thus available for use as a root file system, or for any other file system. A primary advantage of using the auto-configuration is that RAID components become more independent of the disks they reside on. For example, SCSI ID's can change, but auto-configured sets will always be configured correctly, even if the SCSI ID's of the component disks have become scrambled.

Having a system's root file system (/) on a RAID set is also allowed, with the ‘a’ partition of such a RAID set being used for /. To use raid0a as the root file system, simply use:

# raidctl -A root raid0

To return raid0 to be just an auto-configuring set simply use the -A yes arguments.

Note that kernels can't be directly read from a RAID component. To support the root file system on RAID sets, some mechanism must be used to get a kernel booting. For example, a small partition containing only the secondary boot-blocks and an alternate kernel (or two) could be used. Once a kernel is booting however, and an auto-configured RAID set is found that is eligible to be root, then that RAID set will be auto-configured and its ‘a’ partition (aka raid[0..n]a) will be used as the root file system. If two or more RAID sets claim to be root devices, then the user will be prompted to select the root device. At this time, RAID 0, 1, 4, and 5 sets are all supported as root devices.

A typical RAID 1 setup with root on RAID might be as follows:

1. wd0a - a small partition, which contains a complete, bootable, basic OpenBSD installation.
2. wd1a - also contains a complete, bootable, basic OpenBSD installation.
3. wd0e and wd1e - a RAID 1 set, raid0, used for the root file system.
4. wd0f and wd1f - a RAID 1 set, raid1, which will be used only for swap space.
5. wd0g and wd1g - a RAID 1 set, raid2, used for /usr, /home, or other data, if desired.
6. wd0h and wd1h - a RAID 1 set, raid3, if desired.

RAID sets raid0, raid1, and raid2 are all marked as auto-configurable. raid0 is marked as being a root-able raid. When new kernels are installed, the kernel is not only copied to /, but also to wd0a and wd1a. The kernel on wd0a is required, since that is the kernel the system boots from. The kernel on wd1a is also required, since that will be the kernel used should wd0 fail. The important point here is to have redundant copies of the kernel available, in the event that one of the drives fail.

There is no requirement that the root file system be on the same disk as the kernel. For example, obtaining the kernel from wd0a, and using sd0e and sd1e for raid0, and the root file system, is fine. It is critical, however, that there be multiple kernels available, in the event of media failure.

Multi-layered RAID devices (such as a RAID 0 set made up of RAID 1 sets) are not supported as root devices or auto-configurable devices at this point. (Multi-layered RAID devices are supported in general, however, as mentioned earlier.) Note that in order to enable component auto-detection and auto-configuration of RAID devices, the line:

option	RAID_AUTOCONFIG

must be in the kernel configuration file. See raid(4) for more details.

Unconfiguration

The final operation performed by raidctl is to unconfigure a raid(4) device. This is accomplished via a simple:

at which point the device is ready to be reconfigured.

Performance Tuning

Selection of the various parameter values which result in the best performance can be quite tricky, and often requires a bit of trial-and-error to get those values most appropriate for a given system. A whole range of factors come into play, including:

1. Types of components (e.g. SCSI vs. IDE) and their bandwidth
2. Types of controller cards and their bandwidth
3. Distribution of components among controllers
4. I/O bandwidth
5. File system access patterns
6. CPU speed

As with most performance tuning, benchmarking under real-life loads may be the only way to measure expected performance. Understanding some of the underlying technology is also useful in tuning. The goal of this section is to provide pointers to those parameters which may make significant differences in performance.

For a RAID 1 set, a SectPerSU value of 64 or 128 is typically sufficient. Since data in a RAID 1 set is arranged in a linear fashion on each component, selecting an appropriate stripe size is somewhat less critical than it is for a RAID 5 set. However: a stripe size that is too small will cause large I/Os to be broken up into a number of smaller ones, hurting performance. At the same time, a large stripe size may cause problems with concurrent accesses to stripes, which may also affect performance. Thus values in the range of 32 to 128 are often the most effective.

Tuning RAID 5 sets is trickier. In the best case, I/O is presented to the RAID set one stripe at a time. Since the entire stripe is available at the beginning of the I/O, the parity of that stripe can be calculated before the stripe is written, and then the stripe data and parity can be written in parallel. When the amount of data being written is less than a full stripe worth, the ‘small write’ problem occurs. Since a ‘small write’ means only a portion of the stripe on the components is going to change, the data (and parity) on the components must be updated slightly differently. First, the ‘old parity’ and ‘old data’ must be read from the components. Then the new parity is constructed, using the new data to be written, and the old data and old parity. Finally, the new data and new parity are written. All this extra data shuffling results in a serious loss of performance, and is typically 2 to 4 times slower than a full stripe write (or read). To combat this problem in the real world, it may be useful to ensure that stripe sizes are small enough that a ‘large I/O’ from the system will use exactly one large stripe write. As is seen later, there are some file system dependencies which may come into play here as well.

Since the size of a ‘large I/O’ is often (currently) only 32K or 64K, on a 5-drive RAID 5 set it may be desirable to select a SectPerSU value of 16 blocks (8K) or 32 blocks (16K). Since there are 4 data sectors per stripe, the maximum data per stripe is 64 blocks (32K) or 128 blocks (64K). Again, empirical measurement will provide the best indicators of which values will yield better performance.

The parameters used for the file system are also critical to good performance. For newfs(8), for example, increasing the block size to 32K or 64K may improve performance dramatically. Also, changing the cylinders-per-group parameter from 16 to 32 or higher is often not only necessary for larger file systems, but may also have positive performance implications.

Summary

Despite the length of this man-page, configuring a RAID set is a relatively straight-forward process. All that needs to be done is the following steps:

1. Use disklabel(8) to create the components (of type RAID).
2. Construct a RAID configuration file: e.g. ‘raid0.conf’
3. Configure the RAID set with:

   # raidctl -C raid0.conf raid0

4. Initialize the component labels with:

   # raidctl -I 123456 raid0

5. Initialize other important parts of the set with:

   # raidctl -i raid0

6. Get the default label for the RAID set:

   # disklabel raid0 > /tmp/label

7. Edit the label:

   # vi /tmp/label

8. Put the new label on the RAID set:

   # disklabel -R raid0 /tmp/label

9. Create the file system:

   # newfs /dev/rraid0e

10. Mount the file system:

    # mount /dev/raid0e /mnt

11. Use:

    # raidctl -c raid0.conf raid0

    to re-configure the RAID set the next time it is needed, or put raid0.conf into /etc where it will automatically be started by the /etc/rc scripts.