 |
Index for
Section 5 |
|
 |
Alphabetical
listing for S |
|
 |
Bottom of
page |
|
sys_attrs_advfs(5)
NAME
sys_attrs_advfs - advfs subsystem attributes
DESCRIPTION
This reference page lists and describes attributes for the Advanced File
System (advfs) kernel subsystem. Refer to the sys_attrs(5) reference page
for an introduction to the topic of kernel subsystem attributes.
An asterisk (*) preceding the name of an attribute in the following list
means that the attribute can be configured at run time. Values of other
attributes cannot be changed without rebooting the system.
* AdvfsAccessMaxPercent
The maximum percentage of the malloc pool (pageable memory) that can be
possibly be allocated for AdvFS access structures.
Default value: 25 (percent)
Minimum value: 5 (percent)
Maximum value: 95 (percent)
AdvfsAccessMaxPercent is a high-water mark of possible memory use for
access structures; however, the values of the AdvfsMinFreeAccessPercent
and AdvfsMaxFreeAccessPercent attributes control allocation and
deallocation of access structures. If AdvFS cannot allocate any more
access structures because of the memory limit imposed by
AdvfsAccessMaxPercent, AdvFS posts an event and writes a message in the
following form to the console:
Could not create new AdvFS access structure;
already at AdvFS memory limit as specified by
AdvfsAccessMaxPercent value of percent.
At this point, the system administrator can either kill processes to
cause files to be closed or set AdvfsAccessMaxPercent to a higher
value.
AdvfsCacheHashSize
The number of buckets in the hash table for buffers in the AdvFS buffer
cache.
Default value: The size of the AdvFS buffer cache divided by 10 and
rounded up to the next power of 2 or 8192, whichever is less. (The size
of the AdvFS buffer cache is set by the AdvfsCacheMaxPercent
attribute.)
Minimum value: 1024 (entries)
Maximum value: 65,536 (entries)
By decreasing AdvfsCacheHashSize, you save a small amount of kernel
space at the expense of longer hash queue search time.
AdvfsCacheMaxPercent
The percentage of system memory that is allocated to the AdvFS buffer
cache, which stores file data.
Default value: 7 (percent)
Minimum value: 1 (percent)
Maximum value: 30 (percent)
Increasing this percentage can help improve disk I/O only if there is
high reuse of cached data by your application workload. Decreasing this
percentage can improve performance if your application workload does
not reuse AdvFS data or if your system has more than 2 GB of memory.
* AdvfsDomainPanicLevel
One of the following four values that control what AdvFS does in the
event of a domain panic. The live_dump() function mentioned in these
descriptions is a system call used only inside the kernel. This system
call creates vmunix and vmcore files in the /var/adm/crash directory.
In other words, values that enable the live_dump() call allow the
creation of dump files without calling panic() to crash the system.
0 Do not call live_dump() for a domain panic in either an
unmounted or mounted domain.
1 (the default)
Call live_dump() only if the domain panic occurred in a domain
that is mounted.
2 Call live_dump() if a domain panic occurred in either a mounted
or unmounted domain.
3 Promote the domain panic to a system panic. This causes the
system to crash, which yields a crash dump when the system is
rebooted.
* AdvfsFavorBlockingQueue
The value that controls the priority for moving synchronous or
asynchronous I/O requests from the consolidation queue to the device
queue. When the value is 1, AdvFS first flushes synchronous I/O to
disk. When the value is 0 (zero), asynchronous I/O is flushed to disk,
regardless of synchronous I/O.
Default value: 1
Do not modify the default setting for this attribute unless instructed
to do so by support personnel or by patch kit documentation.
* AdvfsMaxDevQLen
The highest number of I/O requests that can be on the AdvFS device
queue before that queue is limited to synchronous requests. The AdvFS
device queue accepts both asynchronous I/O requests and synchronous I/O
requests. When the number of requests on the device queue exceeds the
value of this attribute, only synchronous requests are accepted onto
the queue.
Default value: 24 (requests)
Minimum value: 0 (requests)
Maximum value: 65536 (requests)
The default value should be appropriate for most configurations.
However, you may need to modify the default value for systems with very
fast or very slow devices and adapters. One guideline is to specify a
value for the AdvfsMaxDevQLen attribute that is less than or equal to
the average number of I/O operations that can be performed in 0.5
seconds. If you do not want to limit the size of the device queue, set
the value of the AdvfsMaxDevQLen parameter to 0 (zero).
* AdvfsMaxFragGrps
The threshold number of fragment groups on the free list that starts
fragment group deallocation from the list. Deallocation starts when the
number of fragment groups on the free list is higher than this value.
Default value: 48 (groups)
Minimum value: 4 (groups)
Maximum value: 8,388,608 (groups)
* AdvfsMaxFreeAccessPercent
The maximum percentage of all access structures that can possibly be on
the free list at one time. For example, if AdvfsMaxFreeAccessPercent is
set to 80, access structures are deallocated from the free list when it
contains 80 percent or more of the access structures currently
allocated. The amount of pageable memory that the system is allowed to
allocate to AdvFS access structures is set by the AdvfsAccessMaxPercent
attribute.
Default value: 80 (percent)
Minimum value: 5 (percent)
Maximum value: 95 (percent)
Increasing this attribute value causes the system to retain access
structures on the free list for a longer period of time, which may
improve system performance if your application workload opens and
reuses many files. If your application workload does not open and reuse
many files, decreasing this attribute value may improve performance by
freeing memory otherwise allocated to the free list.
Note that there are actually two requirements that must be met before
the system deallocates access structures from the free list.
AdvfsMaxFreeAccessPercent specifies one requirement. The other is that
the free list must contain more than twice the number of access
structures specified by the AdvfsMinFreeAccess attribute.
* AdvfsMinFragGrps
The threshold number of fragment groups on the free list that stops
fragment group deallocation from the list. Deallocation stops when the
number of fragment groups on the free list is less than this value.
Default value: 16 (groups)
Minimum value: 3 (groups)
Maximum value: 8,388,607 (groups)
* AdvfsMinFreeAccess
The minimum number of AdvFS access structures that can be on the access
structure free list at one time. As AdvFS files are opened, the system
allocates access structures from the free list. When the number of
access structures on the free list falls below AdvfsMinFreeAccess, the
system allocates additional access structures to the free list. When
the number of access structures on the free list is more than twice the
number of AdvfsMinFreeAccess and exceeds the value of the
AdvfsMaxFreeAccessPercent attribute, the system deallocates access
structures from the free list.
Default value: 128 (access structures)
Minimum value: 1 (access structures)
Maximum value: 100,000 (access structures)
If your system workload opens and reuses many files, increasing this
attribute value may improve performance by delaying deallocation of
access structures from the free list. If your system workload does not
open and reuse many files, decreasing this attribute value may improve
performance by freeing memory otherwise allocated to the free list.
* AdvfsReadyQLim
The number of 512-byte blocks that can be on the readylazy queue before
AdvFS starts moving I/O requests to the device queue.
Default value: 16 K (blocks)
Minimum value: 0, which disables buffering on the readylazy queue
Maximum value: 32 K (blocks)
* AdvfsSyncMmapPages
A value that controls whether modified (dirty) mmapped pages are
flushed to disk during a sync() system call. If the value is 1, the
dirty mmapped pages are asynchronously written to disk. If the value
is 0 (zero), dirty mmapped pages are not written to disk during a sync
system call.
Default value: 1
SEE ALSO
sys_attrs(5)
AdvFS Administration
System Configuration and Tuning
|
Index for
Section 5 |
|
|
Alphabetical
listing for S |
|
 |
Top of
page |
|