View HP-UX Kernel Parameter commands

View HP-UX Kernel Parameter commands

Sysdef
Kmtune-l

For reference only:

Acctresume and acctsuspend
Used only when HP-UX statistics are enabled. These variables are the percentage of the file system where the log file is located (/var/adm by default. In the file system
When the space is reduced to the percentage (absolute percentage) specified by acctsuspend, the statistics are terminated. The statistics can be restored only when the percentage allocated to acctresume is reached.
.
For example:
Assign 0 to acctsuspend (assuming the default file system value )-
When the free space is lower than minfree (10% by default, the file system displays
100%), the statistics will be terminated. If acctresume is 80, when the file system's utilization rate drops to 80% (BDF is displayed), the system will be enabled again.
. After statistics are re-enabled, "Accounting resumed" information is generated.
For more information, see the statistical White Paper and statistical help information listed in/usr/share/doc/doc_map.txt.

Bufpages
This value was previously used to define the amount of physical memory allocated by the High-Speed Buffer Used in file system io (in 4096 bytes page ).
Previous HP-UX versions typically used 10% of physical memory for this task, but recent versions have implemented dynamic memory allocation. In version 10. X, if
Bufpages is a non-zero value, which becomes the maximum value of the pages available in the high-speed buffer zone. It actually becomes a limit. Although it is rarely used, it will not exceed this value. In version 10. X, bufpages is often set to 0, which indicates the request for dynamic high-speed buffer, dbc_min_pct
And the dbc_max_pct parameter set the minimum and maximum percentage of available memory allowed by a high-speed buffer.
In version 9. x, the memory of the high-speed buffer zone is determined by the bufpages variable. If/etc/CONF/dfile
This variable is missing from (700 series) or/etc/CONF/GEN/s800 (800 series), and the high-speed buffer zone is set to 10% of the available memory; otherwise, this value is set to the number of pages (4096 bytes.

Create_fastlinks
Allows you to create high-speed symbolic links in the HFS file system. The version notes should contain additional information. In essence, high-speed symbolic links reduce disk space
Block access to slightly reduce disk Io.
* Note: This variable is not found in 10.0 systems earlier than 800 or 9.0 systems earlier than 700.
Dbc_max_pct and dbc_min_pct
These two variables define the buffer file system page (also called high-speed buffer) available memory percentage range. An appropriate value can generally produce the following results:
-Read buffer hit rate lower than or equal to 95%-write buffer hit rate lower than or equal to 70%
Sar-B 5 5 (% rcache and % wcache respectively) can be used to monitor this value. It may reduce read buffer hit.
Another symptom of allocating too much memory for high-speed buffering may be occasional or intermittent pauses that cannot be explained during the user response time. The default value of dbc_min_pct is 5,
The default value of dbc_max_pct is 50. In many cases, we recommend that you allocate MB or less memory for the high-speed buffer. Dbc_max_pct is the one on the machine
It mainly reduces the object, in which we can observe the memory pressure and the pause described just now.

Default_disk_ir
Disk instant report. This variable determines whether the write () System Call waits for data to be actually written to the disk, or only writes data to the disk cache. The default action is disabled.
, Indicates that write () will be written to the disk rather than the disk cache. Additional information is provided in the version notes.

Fs_async
Allows write operations on asynchronous file systems. Although this may improve the file system performance in some cases, the file system has a larger corrupt window. We recommend that you use other
To ensure data integrity. Please be careful to change this variable, which is only related to the file system and not to the original disk device or the original lvols.

Maxdsiz, maxssiz, and maxtsiz
These parameters are three basic components of all HP-UX user processes (data, stack, and text segments ). The maximum size of the data segment is 1.9 GB, but this value
Exec_magic is required to compile and link executable code (for details, see LD (1) Help option-N and version notes ). Exec_magic removes several process Sharing
The same memory image capability of the Code. Without exec_magic, maxdsiz has a maximum of approximately 966 MB. If the data segment (maxdsiz) of each process exceeds

After this value, the binary code must be relinked at least with LD (1) or recompiled according to the settings of exec_magic. For more information, see LD (1) and current
Version description.

Note: Version 10.10 enhances exec_magic so that text images can be shared among multiple processes. For details, see help information and version notes.

Maxfiles & maxfiles_lim
This variable is related to the software and hardware limitation of the maximum number of files that a process can open at a time. The default value of maxfiles is 60.

And stderr. Therefore, you can open only 57 more files. Non-root system process (UID> 0) can add the soft limit to the maxfiles_lim hard limit, but only can pass
Application setrlimit (2) system call. Otherwise, the only way to add soft limit is to change the value of maxfiles.

Maxswapchunks
This value is related to the number of swapchunks (swchunk) that can be used for allocation. Because swap space is allocated with swapchunks, this variable must be large enough to access

For some SWAP spaces, the default value 256 allows the allocation of approximately 525 MB swap space. If this value is too small for the configured swap space, the undefined swap cannot be obtained.
Usually marked as retained and inaccessible, thus limiting the swap space and Total virtual memory. The cost of this variable is not important for memory. Set
Setting maxswapchunks to 1024 and 2048 enables the kernel to address swap areas that are much larger than most configurations.
Maxuprc
This value determines the number of processes that a user identity can generate. This is a limiting factor for each uid, regardless of the number of Logon Processes. Large maxuprc
Value can generate an environment in which user processes Out Of Control consume a large amount of processor time due to the large number of messy processes. It may also affect
Other inner variables reduce the machine's throughput. We recommend that you be careful when adding the value of this variable.
Maxusers
Contrary to some people's views, this variable has nothing to do with the number of users allowed to log on to the HP-UX, the main function of this variable is similar to maxuprc-a larger number of cores
The input variable of the company generally does not need to be changed. It is wise to focus on the specific core resources that need to be modified.
Maxvgs

This parameter indicates the maximum number of volumes that can be configured under the current Kernel configuration. The default value is 10. If vgcreate returns information that cannot open the VG group file, check
The serial number of the new volume group file. If the volume group sub-number is 0x0a0000 or greater, the default value of maxvgs will return an error.

Nbuf

This parameter is out of date. Currently, this value is mainly used on the 300/400 series platform for configuring high-speed buffers. It is used with bufpages-every two
One bufpages and one nbuf. If it is used on machines of the 700 or 800 series, an error occurs in/etc/dmesg or syslog, indicating that the number of nbufs instances has been tuned

. For use of HP-UX
We do not recommend that you use this parameter instead of bufpages, dbc_max_pct, or dbc_min_pct.
. If nbuf is completely removed from the list of core variables, future use may cause errors.

Ncallout

The maximum number of timeouts that can be scheduled by the kernel at any time. The higher the number of processes, the higher the possibility of multiple timeout for each process, the higher the configuration of this value. Default Value
Is the formula (16 + nproc ).

Netmemmax

IP data packets are resegmented and combined into memory. The value-1 indicates that there is no limit on the physical memory, while the value 0 indicates that the limit is 10%. a value greater than 0 indicates that it may be close to the last page boundary.
The amount of physical memory. This memory is not retained and can only be allocated as needed. The adjustment of this variable is to flatten between the user memory and the kernel network memory.
Balance, you may need to do some experiments to find the optimal value.

Nfile

Number of files opened by all processes running on the system. Although each logon item is relatively small, there will still be some kernel overhead when managing this table. This
In addition, each time a file is opened, a logon entry is consumed in nfile, even if the file has been opened by another process. When the nfile logon item is used up
A console and/or syslog error message is displayed, which explicitly states "file table full ".

Nflocks

Number of File locks in the system. Similar to nfile, each lock generates a logon entry, and the memory cost is very small. Each nflock logon entry uses a lot of memory.
Less.

Ninode

This notorious variable is too large to carry too much burden on the processor (especially on machines with multiple CPUs ). No valid statistics for this table, because
It is difficult to monitor the actual consumption. The best way is not to increase the value unless you receive the console/syslog information and specify "inode table is
"Full". Otherwise, after starting for a period of time, it seems that the table is almost or completely filled.

Unlike nfile, each time a file is opened, only one logon entry is consumed in ninode. Generally, there is no direct association between nfile and ninode. In addition, too high
The value may actually cause the high-availability cluster to experience network timeout, often when the backup routine is started. Few systems actually need more than 5,000 logins
. If this variable is large, the initial wait time for listing a logon item is very large, so that the file can be opened quickly first.

Because there is no valid statistics, the only method to determine what content in this table is sequential search, which consumes a lot of processing time. When the processor "goes
When this table is used, few other activities are executed. It is recommended that the table be a little smaller. The value of this table is increased only when the inode table is full.
Increase by 10-20%.

Dnlc

This parameter is not directly adjustable, but the operation is similar to ninode. It is the Directory Name Search Buffer, used for the directory name rather than the file, is a parameter affected by maxusers
I. If the kernel scheme fails, this variable generally uses the value of ninode. When this value is set to a large value, the result is similar to that of ninode. Check the most
The simple method is to use the glance-T command and then check the second page, which contains ninode and dnlc. SAR cannot check this value. Similarly, long text is not retained in this table.
The part name (more than 14 characters) must be fully interpreted each time the table is referenced.

No_lvm_disks

This variable notifies the kernel to check the LVM disk. When this variable is set to 1, no check is performed.

Nproc

This is another variable affected by the messy settings of maxusers/maxuprc. It is related to the number of processes in the system and is often running PS
-Ef, or glance/GPM
And similar commands. This value should generally be 10-20% larger than the maximum number of processes reserved for an unexpected process increase.

Shmmax

This is the maximum byte value of each shared memory segment. In most cases, the maximum value is 1 GB. Phkl_8327 or later products will allow all
The cumulative sum of shared segments. The maximum value is 2.75 GB, while each segment is still limited to 1 GB.

Swapmem_on

This value allows the process to release/restore the allocated memory. It can be observed under the swapinfo command that the display is memory, although the default priority of pseudo-swap is 12,
However, when the memory starts, it is used for invalid processes. The main result is that I/O is reduced, because invalid processes do not have to be migrated to the disk swap area or file system.

If the memory pressure increases and more memory pages are required, invalid processes will be moved to the swap area. Use swapinfo-ATM to observe the swap area in use
No harm. You can use the vmstat command to observe the memory pressure. If swapinfo shows that the number of rows reserved by the device decreases and the number of PCT used columns increases,
You can also verify this.

Timeslice

This value is related to the number of time-segment signals that can be consumed by each process before the priority is reduced and the processor is temporarily lost. Generally, changing this parameter interests us.
In most cases, modification is not recommended or necessary.