sys_attrs_cfs(5)sys_attrs_cfs(5)NAMEsys_attrs_cfs - cfs subsystem attributes
DESCRIPTION
This reference page lists and describes attributes for the cluster
file system (cfs) kernel subsystem. Refer to the sys_attrs(5) reference
page for an introduction to the topic of kernel subsystem attributes.
Size of the hash table used to hold CFS-type vnodes (cnodes).
Set this attribute relative to the maximum number of vnodes
(files) allowed in the system. You can learn this number with
the following command:
sysconfig -q vfs max-vnodes
Supported operations: Query; Configure at boot time
Default value: If a value is not explicitly set in /etc/syscon‐
figtab, a value is calculated by dividing the value of max-
vnodes by 12. If the result is even, 1 is added to it to make
cnode_hash_size odd. This is done to improve hashing perfor‐
mance. The result is 12 cnodes on each hash chain (on average).
This is appropriate for most cases.
Minimum value: 64
Maximum value: 16384
The percent of total space saved for block reservation. The
default value is 50MB or blkrsrv_save_percent (10) percent of
the total free space, whichever is smaller depending on the size
of the domain. If free space reaches this number, all write
operations are sent to the CFS server.
Supported operations: Query; Configure at boot time
Default value: 10
Minimum value: 5
Maximum value: 100
Enables and disables the debugging of certain sections of CFS
client and server code.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 768
Minimum value: -9223372036854775808
Maximum value: 9223372036854775807
Specifies which code sections are to be debugged and how the
debug messages are output.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 0
Minimum value: -9223372036854775808
Maximum value: 9223372036854775807
Maximum size of the CFS event trace log.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Default value: 81920
Minimum value: 8192
Maximum value: 1638400
If set to 1, prepend timestamp, command name and thread ID to
messages written to the CFS event trace log.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 1
Minimum value: -2147483648
Maximum value: 2147483647
Number of seconds before a CFS thread that is hung waiting for a
token is timed out.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 4
Minimum value: -9223372036854775808
Maximum value: 9223372036854775807
When set to 1, panics this member on certain CFS failures. Oth‐
erwise, the failures are logged to the CFS event trace log.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 0
Minimum value: -9223372036854775808
Maximum value: 9223372036854775807
When set to 1, turns on CFS client and server token logging.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 3
Minimum value: -2147483648
Maximum value: 2147483647
Size in bytes of server token log.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 2000
Minimum value: -2147483648
Maximum value: 2147483647
Size in bytes of client token log.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 1000
Minimum value: -2147483648
Maximum value: 2147483647
Percentage of memory that can be used by the token data struc‐
tures on a CFS server.
The svrcfstok_max_percent attribute sets an upper limit on the
amount of memory used by a CFS server to track vnode caching on
clients.
Without this attribute, the only limit is the number of vnodes
that can be created at CFS client nodes. If the number of CFS
clients is large and a single cluster member serves many file
systems, the server could run out of memory and hang.
Supported operations: Query; Configure at boot time
Default value: 25
Minimum value: 5
Maximum value: 50
The default value is well suited for most situations. For more
information, see the TruCluster Server Administration manual.
On large configurations, the synchronization of file systems can
take considerable time, slowing failover. However, if no syn‐
chronization is done, data not flushed to disk is lost if the
data originated from an application on the panicking member.
Data originating from an application on a remote, surviving mem‐
ber still exist on that member and is rescheduled when a new
server for the file system is selected.
The default, 0 (zero), allows writing buffers when the system
panic.
When no_sync_on_panic is 1, the normal process of attempting to
write modified buffers back to disk when the system panics is
disabled. File corruption can result.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 0
Minimum value: 0
Maximum value: 1
Maximum number of cached objects on the token hold freelist.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time
Default value: 1024
Minimum value: -2147483648
Maximum value: 2147483647
Maximum number of cached objects on the token request freelist.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time
Default value: 512
Minimum value: -2147483648
Maximum value: 2147483647
Maximum number of cached objects on the token revoke freelist.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Supported operations: Query; Configure at boot time
Default value: 256
Minimum value: -2147483648
Maximum value: 2147483647
The bread attribute enables an optimization that eliminates a
data copy for NFS read requests. There should be no reason to
change this attribute from its default value.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 1
Minimum value: 0
Maximum value: 1
Controls the number of threads on the member that do read-aheads
and write-behinds for applications that perform sequential reads
and writes. The read aheads and write behinds are performed only
on CFS clients (not the CFS server).
Setting this attribute may be useful in an environment in which
sequential read and write performance of large files is impor‐
tant. By default, CFS reads in 64 KB chunks. A read-ahead
thread is useful when a file has multiple 64 KB chunks that are
read sequentially. Write behind threads are useful to avoid
blocking threads that are writing, and to avoid blocking threads
that are reclaiming free dirty CFS client pages in the buffer
cache.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 32
Minimum value: 0
Maximum value: 128
A file in a Hierarchical Storage Management (HSM) file system
managed by a UniTree application is considered by CFS to be busy
if the file was accessed in the past cfs_hsm_min_access_time
seconds.
The cfs_hsm_min_access_time attribute must have the same value
on every member of the cluster. If you modify the value of the
attribute, you must do so on all members.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 300
Minimum value: 0
Maximum value: 4294967295
After an initial mount of a Hierarchical Storage Management
(HSM) file system that is to be managed by a UniTree applica‐
tion, and during a failover or relocation of the CFS server for
an HSM file system, the HSM file system is unavailable until the
UniTree application starts. The cfs_hsm_timeout attribute spec‐
ifies the number of seconds that CFS holds on to an access
request to the HSM managed file system. This delay provides time
for the UniTree application to start.
If cfs_hsm_timeout is set to -1 then timeout never occurs, and
CFS holds on to the access request indefinitely.
The attribute must have the same value on every member of the
cluster. If you modify the value of the attribute, you must do
so on all members.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 15
Minimum value: -1
Maximum value: 2147483647
The cfsiosize attribute controls the amount of data that is pro‐
cessed per pass through the client side I/O code path. The value
is in kilobytes. Making cfsiosize large results in more data
being proceesed per pass. Typically, this yields better perfor‐
mance.
To set the I/O size for a specific file system, overriding the
effect of cfsiosize on that file system, use the cfsmgr command
to set the FSBSIZE attribute for the file system.
See cfsmgr(8) for more information.
Supported operations: Query; Configure at boot time; Reconfigure
at run time
Default value: 65536
Minimum value: 8192
Maximum value: 131072
The quota_excess_blocks attribute controls the amount of dirst
data that a given user or group is allowed to cache when user or
group quotas are enabled.
The quota_excess_blocks value is expressed in units of 1024-byte
blocks and the default value is 1024, which represents one
megabyte of disk space. The value of quota_excess_blocks does
not have to be the same on all cluster members. You might use a
larger quota_excess_blocks value on cluster members on which you
expect most of the data to be generated, and accept the default
value for quota_excess_blocks on other cluster members. Use the
sysconfigdb command to make changes.
Performance for a given user or group can be affected by
quota_excess_blocks. If this value is set too low, CFS will not
be able to use the cache efficiently. Setting
quota_excess_blocks to less than 64K will have a severe perfor‐
mance impact. Conversely, setting quota_excess_blocks too high
increases the actual amount of disk space a user or group can
consume above their quota. We recommend accepting the
quota_excess_blocks default of one megabyte, or increasing it as
much as is considered practical given its effect of raising the
potential upper limit on disk block usage. When determining how
to set this value, consider that the worst-case upper boundary
is determined as follows:
(admin specified hard limit) +
(sum of "quota_excess_blocks" on each client node)
Supported operations: Query; Configure at boot time; Reconfig‐
ure at run time.
Default value: 1024, in units of 1024-byte blocks
Minimum value: 0
Maximum value: 18446744073709551615
SEE ALSOsys_attrs(5), cfsmgr(8)
TruCluster Server Cluster Administration
sys_attrs_cfs(5)