Skip Navigation Links | |
Exit Print View | |
Sun QFS File System 5.3 Configuration and Administration Guide Sun QFS and Sun Storage Archive Manager 5.3 Information Library |
2. About the Master Configuration File
4. Configuring the File System
5. Configuring a Shared File System
6. Administering File System Quotas
7. Advanced File System Topics
9. Configuring WORM-FS File Systems
11. Using QFS File Systems with SANergy (SAN-QFS)
12. Mount Options in a Shared File System
13. Using the samu Operator Utility
samu Operator Utility at a Glance
Overview of the samu Operator Utility
(c) - Device Configuration Display
(d) - Daemon Trace Controls Display
(F) - Optical Disc Label Display
(J) - Preview Shared Memory Display
(K) - Kernel Statistics Display
(m) - Mass Storage Status Display
(N) - File System Parameters Display
(o) - Optical Disc Status Display
(p) - Removable Media Load Requests Display
(r) - Removable Media Status Display
(R) - Sun SAM-Remote Information Display
(t) - Tape Drive Status Display
(v) - Automated Library Catalog Display
Removable Media Device Display Status Codes
File System Display Status Codes
States of an Operator Display Device
How to Change a Drive State From down to on
File System Commands: I/O Management
force_nfs_async and noforce_nfs_async Commands
sw_raid and nosw_raid Commands
File System Commands: Direct I/O Management
dio_rd_form_min and dio_wr_form_min Commands
dio_rd_ill_min and dio_wr_ill_min Commands
dio_rd_consec and dio_wr_consec Commands
dio_szero and nodio_szero Commands
forcedirectio and noforcedirectio Commands
File System Commands: Sun QFS Shared File Systems
mh_write and nomh_write Commands
minallocsz and maxallocsz Commands
rdlease, wrlease, and aplease Commands
File System Commands: Miscellaneous
The following topics describe the operator commands that you can issue from the samu operator utility's command interface. You can issue the commands from any display.
The operator commands types are described in the following sections:
Device Commands
File System Commands: I/O Management
File System Commands: Direct I/O Management
File System Commands: Sun QFS Shared File Systems
File System Commands: Miscellaneous
Miscellaneous Commands
If you want to issue any operator commands from the Oracle Solaris OS command line, you must use them as arguments to the samcmd command. For more information about the samcmd command, see samcmd(1M) in Sun QFS and Sun Storage Archive Manager 5.3 Reference Manual.
The following table shows the device commands and their actions.
Table 13-7 Device Command Actions
|
All these commands use the following syntax:
:command eq
For eq, specify the equipment Nnumber of the device.
The commands described in this section enable you to manage I/O characteristics dynamically.
The flush_behind command sets the maximum flush_behind value. When set to a value greater than 0, modified pages that are being written sequentially are written to disk asynchronously to help the Oracle Solaris kernel layer keep the pages clean. By default, the maximum value is 0, which disables flush_behind.
:flush_behind eq value
For value, specify an integer number of kilobytes such that 0≤ value ≤ 8192.
For eq, specify the equipment number for the file system.
These commands enable you to control whether the file system caches NFS data written to the server even if NFS has requested that the data be written synchronously through to disk. The force_nfs_async command caches NFS data.
The force_nfs_async command is effective only if the file system is mounted as an NFS server and only if the clients are mounted with the noac NFS mount option. For more information about mounting an NFS file system, see the mount_nfs(1M) man page.
Caution - The force_nfs_async option violates NFS protocols. Use this command with caution. In the event of a server interruption, data can be lost. Data is cached on the NFS server and cannot be seen immediately by all the clients if there are multiple NFS servers. Multiple NFS servers can be enabled within the Sun QFS shared file system. For more information about the Sun QFS shared file system, see Chapter 5, Configuring a Shared File System. |
The noforce_nfs_async command, which is the default, synchronously writes data through to disk.
:force_nfs_async eq :noforce_nfs_async eq
For eq, specify the equipment number for the file system.
The readahead command specifies the maximum number of bytes that can be read ahead by the file system. The default contig number is 8 (131072 bytes).
:readahead eq contig
For eq, specify the equipment number for the file system.
For contig, specify units of 1-kilobyte blocks. This value must be an integer such that 1 < contig < 8192. The contig specified is truncated to a multiple of 8 kilobytes.
For example, the following command sets the maximum contiguous block size to 262,144 bytes for the file system defined as equipment number 3:
:readahead 3 256
This value can also be configured in the samfs.cmd file by specifying the readahead directive. For more information, see samfs.cmd(4) in Sun QFS and Sun Storage Archive Manager 5.3 Reference Manual.
These commands specify whether the file system aligns the writebehind buffer. Specify sw_raid if the software RAID feature of a package such as Solaris Volume Manager is also used on this file system. The default setting is nosw_raid.
:sw_raid eq :nosw_raid eq
For eq, specify the equipment number for a file system.
The writebehind command specifies the maximum number of bytes that can be written behind by a file system. The default contig number is 8 (131072 bytes).
:writebehind eq contig
For eq, specify the equipment number for a file system.
For contig, specify units of 1-kilobyte blocks. This value must be an integer such that 1 < contig < 8192.
For example, the following command sets the maximum contiguous block size to 262,144 bytes for the file system defined as equipment number 50:
:writebehind 50 256
This value can also be configured in the samfs.cmd file by specifying the writebehind directive. For more information, see samfs.cmd(4) in Sun QFS and Sun Storage Archive Manager 5.3 Reference Manual.
The wr_throttle command sets the number of outstanding write kilobytes for one file. The default is the number of kilobytes that is 2% of the memory size.
If the percentage cannot be calculated, set the value to 100 MBytes. To determine memory size on Oracle Solaris :
npages = sysconf(_SC_PHYS_PAGES); pagesizeb = sysconf(_SC_PAGESIZE); fprintf(fp, "Memory size:\t\t%.1f MBytes (%.1f M2Bytes, %ld pages, %ld bytes/page)\n", ( / (float)npages * (float)pagesizeb) / 1000000.0, ((float)npages * (float)pagesizeb) / / 1048576.0, npages, pagesizeb);
After you have found the memory size, set the value of wr_throttle to 2% of the total size, as shown in the following table.
|
:wr_throttle eq value
For eq, specify the equipment number for a file system.
For value, specify an integer number of kilobytes. If value=0, there is no limit.
The commands described in this section control I/O on Sun QFS file systems. You can change the type of I/O for an individual file based on I/O size and history. If direct I/O is specified for a file, for example, through the setfa command, these options are ignored and all I/O to regular files is direct, if possible.
These commands refer to both well-aligned and misaligned I/O.
Well-aligned I/O occurs when the file offset falls on a 512-byte boundary and when the length of the I/O transfer is at least 512 bytes.
Misaligned I/O occurs when the file offset does not fall on a 512-byte boundary and the length of the transfer is less than 512 bytes.
For more information about I/O and I/O management, see Chapter 7, Advanced File System Topics.
These commands set the lower limits for well-aligned I/O to the specified number of 1024-byte blocks. Use the dio_rd_form_min command to set the number for read operations, and use the dio_wr_form_min command to set the number for write operations. By default, the minimum number of blocks is 256.
:dio_rd_form_min eq value :dio_wr_form_min eq value
For eq, specify the equipment number for the file system.
For value, specify an integer number of 1024-byte blocks to use for the lower limit. If value=0, automatic I/O switching is disabled.
These commands set the lower limit for misaligned I/O to the specified number of 1024-byte blocks. Use the dio_rd_ill_min command to set the number for read operations, and use the dio_wr_ill_min command to set the number for write operations. By default,the minimum number of blocks is 256.
:dio_rd_ill_min eq value :dio_wr_ill_min eq value
For eq, specify the equipment number for the file system.
For value, specify an integer number of 1024-byte blocks to use for the lower limit. If value=0, automatic I/O switching is disabled.
These commands set the number of consecutive I/O transfers that can occur with a buffer size greater than the specified lower limits. By default, value=0, which means that no default direct read operations occur based on I/O sizes.
:dio_rd_consec eq value :dio_wr_consec eq value
For eq, specify the equipment number for the file system.
For value, specify the number of consecutive I/O transfers with a buffer size greater than the specified lower limit. The specified lower limit is the value of dio_rd_form_min for aligned read operations or dio_rd_ill_min for misaligned read operations.
For more information, see the following sections:
These commands set or clear the direct I/O sparse zeroing mount option.
The dio_szero command causes uninitialized areas of sparse files written with direct I/O to be zeroed when the area is accessed. This behavior makes the sparse file behavior the same as that for paged I/O. By default, sparse files written by direct I/O do not have the uninitialized areas zeroed for performance reasons. The default is nodio_szero.
:dio_szero eq :nodio_szero eq
For eq, specify the equipment number for the file system.
These commands enable you to control whether direct I/O is used as the default I/O mode. By default, the I/O mode is buffered and uses the page cache. The forcedirectio command enables direct I/O for all transfers. The noforcedirectio command restores the default, buffered I/O.
When direct I/O is specified, the system transfers data directly between the users buffer and disk. Use direct I/O only if the file system is used for large, block-aligned, sequential I/O.
:forcedirectio eq :noforcedirectio eq
For eq, specify the equipment number for the file system.
For more information about I/O, see Chapter 7, Advanced File System Topics.
The file system commands described in this section are supported on Sun QFS shared file systems only.
The meta_timeo command sets the time limit for the Sun QFS shared file system metadata cache. The default value is 3. For more information about using this feature, see Retaining Cached Attributes: (meta_timeo Option).
:meta_timeo eq interval
For eq, specify the equipment number of the file system.
For interval, specify time in seconds. After this interval expires, the client host systems obtain a new copy of the metadata information from the metadata server host.
These commands enable or disable multihost read and write operations. For information about this feature, see Enabling Multiple Host Reads and Writes: (mh_write Option).
:mh_write eq :nomh_write eq
For eq, specify the equipment number of the file system.
These commands set the minimum and maximum block allocation size.
:minallocsz eq value :maxallocsz eq value
For eq, specify the equipment number of the file system.
For value, and for more information about this feature, see Tuning Allocation Sizes: (minallocsz and maxallocsz Options).
These commands control the amount of time granted for read, write, and append leases. The default time is 30 seconds. For information about this feature, see Using Leases in a Sun QFS Shared File System: (rdlease, wrlease, and aplease Options).
:rdlease eq interval :wrlease eq interval :aplease eq interval
For eq, specify the equipment number of the file system.
For interval, specify an integer number of seconds, 15 ≤ interval ≤ 600.
The commands described in this section enable you to control leases, allocation sizes, and various other file system characteristics.
These commands set or clear the application binary recovery (ABR) mount option.
Use these commands only in an Oracle RAC environment with Sun QFS asynchronous I/O (AIO). These mount options disable or enable ABR of software mirrors. They apply only to Sun QFS file systems built on Solaris Volume Manager mirrored volumes that support ABR.
:abr eq :noabr eq
For eq, specify the equipment number for the file system.
These commands set or clear the direct mirror reads (DMR) mount option.
Use these commands only in an Oracle Real Application Cluster (RAC) environment with Sun QFS AIO. These mount options disable or enable DMR of software mirrors. They apply only to Sun QFS file systems built on Solaris Volume Manager mirrored volumes that support DMR.
:dmr eq :nodmr eq
For eq, specify the equipment number for the file system.
The invalid command specifies that the file system hold cached attributes for at least the specified number of seconds after a file is modified. You can specify this command only if the file system was mounted originally with the reader mount option. For information about mount options, see mount_samfs(1M) in Sun QFS and Sun Storage Archive Manager 5.3 Reference Manual.
:invalid eq interval
For eq, specify the equipment number for the file system.
For interval, specify the number of seconds to hold the attributes after file modification. For example, assume that interval=30. When you issue an ls command, you might not see a newly created file appear in its output for 30 seconds after it has been created on its writer host.
The mm_stripe command sets the metadata stripe width for the file system to the specified number of 16-kilobyte disk allocation units (DAUs). The default is 1 DAU so that the file system writes one DAU of metadata to one LUN before switching to another LUN.
:mm_stripe eq value
For eq, specify the equipment number of the file system.
For value, specify either 0 or 1. If value is set to 1, which is the default, the file system writes one DAU of metadata to one LUN before switching to another LUN. If value is set to 0, the metadata is round-robined across all available metadata LUNs.
The qwrite and noqwrite commands control the ability to perform simultaneous read and write operations to the same file from different threads. Specify qwrite only if file system users handle multiple simultaneous transactions to the same file. This feature is useful in database applications. The qwrite feature improves I/O performance by queuing multiple requests at the drive level. The qwrite specification is disabled for NFS reads or writes of the file system.
The default setting is noqwrite, so the file system disables simultaneous read and write operations to the same file. This mode is defined by the UNIX vnode interface standard that gives exclusive access to only one writer and forces other writers and readers to wait.
:qwrite eq :noqwrite eq
For eq, specify the equipment number of the file system.
The refresh_at_eof and norefresh_at_eof commands can be used for fast updates to Sun QFS hosts that are mounted with the reader mount option in a multireader file system. This option ensures that the system refreshes the current file size when the read buffer exceeds the end of file. You can use this option,for example, if the writer host system is appending to a file and the reader is issuing tail commands with the -f option. The default is norefresh_at_eof.
:refresh_at_eof eq :norefresh_at_eof eq
For eq, specify the equipment number of the file system.
The suid and nosuid commands control whether running programs are allowed to automatically change their owner IDs. For more information about the implications of using these mount options, see the suid and nosuid mount option descriptions on the mount_ufs(1M) man page and the suid(2) man page.
:suid eq :nosuid eq
For eq, specify the equipment number of the file system.
The stripe command sets the stripe width for the file system to the specified number of disk allocation units (DAUs). The stripe width specifies that value multiplied by the DAU bytes are written to one LUN before switching to the next LUN. You can use the sammkfs -a command to set the DAU size on the file system when it is initialized.
:stripe eq value
For eq, specify the equipment number of the file system.
For value, specify an integer such that 0 < value < 255. If value=0, files are round-robin on each slice. The default value on file systems with an ms equipment type and on file systems with an ma equipment type with no striped group (gXXX) components is as follows:
128 kilobytes/DAU for DAUs < 128 kilobytes
1 for DAUs > 128 kilobytes
By default, value=0 on a Sun QFS shared file system.
By default, value=0 on file systems with an ma equipment type with any striped group (gXXX) components.
The system sets value=0 if mismatched striped groups exist.
For more information about file system types, see Sun QFS File Systems Design Basics and Chapter 4, Configuring the File System.
The sync_meta command determines whether metadata is written to disk every time it changes. If you are using this command on Sun QFS shared file system, also see Specifying the Frequency With Which Metadata Is Written: (sync_meta Option).
:sync_meta eq value
For eq, specify the equipment number of the file system.
For value, specify either 0 or 1, as follows:
For value is 0, metadata is held in a buffer after it changes. For an unshared Sun QFS file system in which higher performance is desired, you can set value to 0. In this case, the system performs a delayed write operation in which metadata is held in a buffer before it is written to disk. This value is the default for unshared file systems and for file systems that are not mounted as multireader file systems.
For value is 1, metadata is written to disk every time it changes. This behavior slows performance, but it increases data consistency. This value is the default for Sun QFS file systems mounted as multireader file systems or as shared file systems. For a Sun QFS shared file system, value must be set to 1 if failover capability is required.
The trace command enables tracing for a file system. The notrace command disables tracing. These global directives affect all operations. For more information about file system tracing, see defaults.conf(4) in Sun QFS and Sun Storage Archive Manager 5.3 Reference Manual.
:trace eq :notrace eq
For eq, specify the equipment number of a file system.
The commands described in this section enable you to control tracing, open access to a disk device, and perform several other miscellaneous tasks.
The clear command clears the specified VSN from the removable media mount requests display. For more information, see (p) - Removable Media Load Requests Display.
:clear vsn :clear vsn index
For vsn, specify the volume to mount. Any process waiting for the VSN mount is aborted.
For index, specify the decimal ordinal of the VSN in the removable media display.
The devlog command sets one or more events to be logged.
:devlog eq :devlog eq option
For eq, specify the equipment number of a device.
For option, specify one or more event types. Possible event types are as follows: all, date, default, detail, err, event, label, mig, module, msg, none, retry, stage, syserr, and time. For information about these options, see defaults.conf(4) in Sun QFS and Sun Storage Archive Manager 5.3 Reference Manual.
If no option is specified, the system does not change the current events being logged for the eq specified.
The diskvols command sets or clears flags in the disk volume dictionary.
:diskvols volume +flag :diskvols volume -flag
For volume, specify the volume in the disk volume dictionary.
For flag, specify one of the five flags in the samu D display. For information about the disk volume dictionary and the flags, see the samu(1M) man page.
The dtrace commands control the DTrace feature for one or more processes. The DTrace commands specify various tracing options.
:dtrace daemon-name on :dtrace daemon-name off :dtrace daemon-name.variable value
For daemon-name, specify the keyword all to affect all processes or a process name. If one of the following process names is specified, the tracing command affects that process only: sam-archiverd,sam-catserverd, sam-fsd, sam-rftd, sam-recycler, sam-sharefsd, and sam-stagerd.
For variable and value, specify one of the following variable and value pairs. The defaults.conf(4) man page contains comprehensive information about these arguments.
filevalue
Specify the name of a file to which trace files can be written. This can be a full path name.
optionsvalue
Specify a space-separated list of trace options.
agevalue
Specify the trace file rotation age.
Note - Do not set an age of two minutes or less. If you do, the rotation never happens.
sizevalue
Specify the size of the trace file at which rotation begins.
The fs command sets the file system to be displayed through the N display.
:fs fsname
For fsname, specify the name of the file system to be examined.
The mount command selects a Sun QFS file system.
:mount mount-point
For mount-point, specify the mount point of a file system.
The open command enables access to the specified disk device. You must issue this command before you can use the read command, disk sector display (S), or file label display (F).
:open eq
For eq, specify the equipment number of a device.
The read command reads the specified sector from the currently opened disk device. You must open the device before it can be read.
:read addr
For addr, specify the hexadecimal sector address.
The refresh command sets the amount of time between samu screen refreshes.
:refresh i
For i, specify a time in seconds.
The snap command sends a snapshot of a display window to file. The default file is snapshots in the current working directory. To aid in problem reporting, take a snapshot of all of the samu utility's displays. Each new snapshot is appended to the snapshots file. The file can be printed, examined using an editor, or faxed to Oracle customer support staff.
:snap :snap filename
For filename, specify the path of a file to receive the display information.
The ! command enables you to run a shell command without leaving the samu operator utility.
:! shell-command
For shell-command, specify a command.