PVFS2 pvfs-2-7-branch: Config File Description
PVFS2 servers are deployed using configuration files that provide information
about the file systems, storage locations and endpoints that each server
manages. For every pvfs2 deployment, there should be a global config file
(fs.conf) shared across all of the pvfs2 servers. When the servers
are started up, a command line parameter (server-alias) indicates what options
are relevant and applicable for a particular server.
This parameter will be used by the server to parse relevant options.
Configuration options in the global config files have the following format:
OptionName OptionValue
An option cannot span more than one line, and only one option can
be specified on each line. The OptionValue should
be formatted based on the option's type:
- Integer - must be an integer value
- String - must be a string without breaks (newlines)
- List - a set of strings separated by commas
Options are grouped together using contexts, and usually
indented within a context for clarity. A context is started
with a context start tag, and ended with a context end tag:
<ContextName>
Option1Name Option1Value
Option2Name Option2Value
</ContextName>
Options are required to be defined within a specified context
or set of contexts.
Sub-contexts can also be specified, and must be defined within
their specified parent context. For example, the Range option is
specified in either the DataHandleRanges or MetaHandleRanges
contexts. Both of
those contexts are specified to be defined in the FileSystem context.
Details of the required context an option or sub-context must be defined in
are given in the Option Details section.
Options and contexts that appear in the top-level (not defined within
another context) are considered to be defined in a special Global
context. Many options are only specified to appear within
the Default context,
which is a context that allows a default value to be specified for certain
options.
The options detailed below each specify their type, the context
where they appear, a default value, and description. The default
value is used if the option is not specified in any of the config files.
Options without default values are required to be defined in the
config file.
Options Grouped by Context
This is the summarized list of all available options, grouped by the
context where they are allowed to appear.
<Defaults> TroveMaxConcurrentIO LogFile LogType EventLogging UnexpectedRequests StorageSpace TCPBufferSend TCPBufferReceive TCPBindSpecific ServerJobBMITimeoutSecs ServerJobFlowTimeoutSecs ClientJobBMITimeoutSecs ClientJobFlowTimeoutSecs ClientRetryLimit ClientRetryDelayMilliSecs PerfUpdateInterval BMIModules FlowModules LogStamp TroveMethod </Defaults> <Security> TrustedPorts TrustedNetwork </Security> <ServerOptions> </ServerOptions> <Aliases> Alias </Aliases> <FileSystem> RootHandle Name ID FlowBufferSizeBytes FlowBuffersPerFlow DefaultNumDFiles SecretKey
<StorageHints> HandleRecycleTimeoutSecs AttrCacheKeywords AttrCacheSize AttrCacheMaxNumElems TroveSyncMeta TroveSyncData DBCacheSizeBytes DBCacheType ImmediateCompletion CoalescingHighWatermark CoalescingLowWatermark TroveMethod </StorageHints> <ExportOptions> </ExportOptions> <DataHandleRanges> Range </DataHandleRanges> <MetaHandleRanges> Range </MetaHandleRanges> <Distribution> Name Param Value </Distribution> </FileSystem> Option Details
| Option: | TrustedPorts |
| Type: | String |
| Contexts: | Security |
| Default Value: | None |
| Description: | Specifies the range of ports in the form of a range of 2 integers
from which the connections are going to be accepted and serviced.
The format of the TrustedPorts option is: TrustedPorts StartPort-EndPort As an example: TrustedPorts 0-65535
|
| Option: | TrustedNetwork |
| Type: | List |
| Contexts: | Security |
| Default Value: | None |
| Description: | Specifies the IP network and netmask in the form of 2 BMI addresses from
which the connections are going to be accepted and serviced.
The format of the TrustedNetwork option is: TrustedNetwork bmi-network-address@bmi-network-mask As an example: TrustedNetwork tcp://192.168.4.0@24
|
| Option: | Alias |
| Type: | List |
| Contexts: | Aliases |
| Default Value: | None |
| Description: | Specifies an alias in the form of a non-whitespace string that
can be used to reference a BMI server address (a HostID). This
allows us to reference individual servers by an alias instead of their
full HostID. The format of the Alias option is: Alias {alias string} {bmi address} As an example: Alias mynode1 tcp://hostname1.clustername1.domainname:12345
|
| Option: | Server |
| Type: | String |
| Contexts: | |
| Default Value: | None |
| Description: | Defines the server alias for the server specific options that
are to be set within the ServerOptions context.
|
| Option: | Range |
| Type: | List |
| Contexts: | MetaHandleRanges DataHandleRanges |
| Default Value: | None |
| Description: | As logical files are created in pvfs, the data files and meta files
that represent them are given filesystem unique handle values. The
user can specify a range of values (or set of ranges)
to be allocated to data files and meta files for a particular server,
using the Range option in the DataHandleRanges and MetaHandleRanges
contexts. Note that in most cases, its easier to let the
pvfs2-genconfig script determine the best ranges to specify. This option specifies a range of handle values that can be used for
a particular pvfs server in a particular context (meta handles
or data handles). The DataHandleRanges and MetaHandleRanges contexts
should contain one or more Range options. The format is: Range {alias} {min value1}-{max value1}[, {min value2}-{max value2},...] Where {alias} is one of the alias strings already specified in the
Aliases context. {min value} and {max value} are positive integer values that specify
the range of possible handles that can be given out for that particular
host. {max value} must be less than 18446744073709551615 (UINT64_MAX). As shown in the specified format, multiple ranges can be specified for
the same alias. The format requires that max value of a given range
is less than the min value of the next one,
i.e. {max value1}<{min value2} Example of a Range option for data handles: Range mynode1 2147483651-4294967297
|
| Option: | RootHandle |
| Type: | String |
| Contexts: | FileSystem |
| Default Value: | None |
| Description: | Specifies the handle value for the root of the Filesystem. This
is a required option in the Filesystem context. The format is: RootHandle {handle value} Where {handle value} is a positive integer no greater than
18446744073709551615 (UIN64_MAX). In general its best to let the pvfs-genconfig script specify a
RootHandle value for the filesystem.
|
| Option: | Name |
| Type: | String |
| Contexts: | FileSystem Distribution |
| Default Value: | None |
| Description: | This option specifies the name of the particular filesystem or
distribution that its defined in. It is a required option in
Filesystem and Distribution contexts.
|
| Option: | ID |
| Type: | Integer |
| Contexts: | FileSystem |
| Default Value: | None |
| Description: | A pvfs server may manage more than one filesystem, and so a
unique identifier is used to represent each one.
This option specifies such an ID (sometimes called a 'collection
id') for the filesystem it is defined in. The ID value can be any positive integer, no greater than
2147483647 (INT32_MAX). It is a required option in the Filesystem
context.
|
| Option: | TroveMaxConcurrentIO |
| Type: | Integer |
| Contexts: | Defaults
|
| Default Value: | 16 |
| Description: | maximum number of AIO operations that Trove will allow to run
concurrently
|
| Option: | LogFile |
| Type: | String |
| Contexts: | Defaults
|
| Default Value: | /tmp/pvfs2-server.log |
| Description: | The gossip interface in pvfs allows users to specify different
levels of logging for the pvfs server. The output of these
different log levels is written to a file, which is specified in
this option. The value of the option must be the path pointing to a
file with valid write permissions. The Logfile option can be
specified for all the pvfs servers in the Defaults context or for
a particular server in the Global context.
|
| Option: | LogType |
| Type: | String |
| Contexts: | Defaults
|
| Default Value: | file |
| Description: | The LogType option can be used to control the destination of log
messages from PVFS2 server. The default value is "file", which causes
all log messages to be written to the file specified by the LogFile
parameter. Another option is "syslog", which causes all log messages
to be written to syslog.
|
| Option: | EventLogging |
| Type: | List |
| Contexts: | Defaults
|
| Default Value: | none |
| Description: | The gossip interface in pvfs allows users to specify different
levels of logging for the pvfs server. This option sets that level for
either all servers (by being defined in the Defaults context) or for
a particular server by defining it in the Global context. Possible
values for event logging are:
| Name | Log output for: |
| acache | Debug the attribute cache. Only useful on the client. | | access | Show server file (metadata) accesses (both modify and read-only). | | access_detail | Show more detailed server file accesses | | access_hostnames | Display the hostnames instead of IP addrs in debug output | | all | Everything | | bstream | | | cancel | Debug the cancel operation | | client | Log client sysint info. This is only useful for the client. | | clientcore | Debug the client core app | | clientcore_timing | Debug the client timing state machines (job timeout, etc.) | | coalesce | Debug the metadata sync coalescing code | | dbpfattrcache | Debug the server-side dbpf attribute cache | | dev | Show the client device events | | distribution | Log/Debug distribution calls | | flow | Log flow calls | | flowproto | Log the flow protocol events, including flowproto_multiqueue | | fsck | Debug the fsck tool | | getattr | Debug the server getattr state machine. | | geteattr | Debug the client and server get ext attributes SM. | | io | Debug the io operation (reads and writes) for both the client and server | | job | Log job info | | keyval | Debug the metadata dbpf keyval functions | | listattr | vectored getattr server state machine | | listeattr | Debug the listeattr operation | | lookup | Debug the client lookup state machine. | | mkdir | Debug the mkdir operation (server only) | | mmaprcache | Debug read-ahead cache events. Only useful on the client. | | msgpair | Debug the msgpair state machine | | ncache | Debug the client name cache. Only useful on the client. | | network | Log network debug info. | | none | No debug output | | open_cache | Debug the server's open file descriptor cache | | permissions | Debug permissions checking on the server | | readdir | Debug the readdir operation (client and server) | | remove | Debug the client remove state macine. | | reqsched | Log request scheduler events | | request | Debug PINT_process_request calls. EXTREMELY verbose! | | server | Log server info, including new operations. | | setattr | Debug the server setattr state machine. | | seteattr | Debug the client and server set ext attributes SM. | | sm | Debug the state machine management code | | storage | Log trove debugging info. Same as 'trove'. | | trove | Log trove debugging info. Same as 'storage'. | | trove_op | Log trove operations. | | varstrip | Debug the varstrip distribution | | verbose | Everything except the periodic events. Useful for debugging | The value of the EventLogging option can be a comma separated list
of the above values. Individual values can also be negated with
a '-'. Examples of possible values are: EventLogging flow,msgpair,io EventLogging -storage EventLogging -flow,-flowproto
|
| Option: | UnexpectedRequests |
| Type: | Integer |
| Contexts: | Defaults
|
| Default Value: | 50 |
| Description: | At startup each pvfs server allocates space for a set number
of incoming requests to prevent the allocation delay at the beginning
of each unexpected request. This parameter specifies the number
of requests for which to allocate space. A default value is set in the Defaults context which will be be used
for all servers.
However, the default value can also be overwritten by setting a separate value
in the ServerOptions context using the Option tag.
|
| Option: | StorageSpace |
| Type: | String |
| Contexts: | Defaults
|
| Default Value: | None |
| Description: | Specifies the local path for the pvfs2 server to use as storage space.
This option specifies the default path for all servers and will appear
in the Defaults context. NOTE: This can be overridden in the tag on a per-server
basis. Look at the "Option" tag for more details
Example: StorageSpace /tmp/pvfs.storage
|
| Option: | TCPBufferSend |
| Type: | Integer |
| Contexts: | Defaults |
| Default Value: | 0 |
| Description: | Current implementations of TCP on most systems use a window
size that is too small for almost all uses of pvfs.
We recommend administators
should consider tuning the linux kernel maximum send and
receive buffer sizes via the /proc settings. The
PSC tcp tuning section for linux has good information
on how to do this. The TCPBufferSend and
TCPBufferReceive options allows setting the tcp window
sizes for the pvfs clients and servers, if using the
system wide settings is unacceptable. The values should be
large enough to hold the full bandwidth delay product (BDP)
of the network. Note that setting these values disables
tcp autotuning. See the
PSC networking options for details.
|
| Option: | TCPBufferReceive |
| Type: | Integer |
| Contexts: | Defaults |
| Default Value: | 0 |
| Description: | See the TCPBufferSend option.
|
| Option: | TCPBindSpecific |
| Type: | String |
| Contexts: | Defaults
|
| Default Value: | no |
| Description: | If enabled, specifies that the server should bind its port only on
the specified address (rather than INADDR_ANY)
|
| Option: | ServerJobBMITimeoutSecs |
| Type: | Integer |
| Contexts: | Defaults
|
| Default Value: | 300 |
| Description: | Specifies the timeout value in seconds for BMI jobs on the server.
|
| Option: | ServerJobFlowTimeoutSecs |
| Type: | Integer |
| Contexts: | Defaults
|
| Default Value: | 300 |
| Description: | Specifies the timeout value in seconds for TROVE jobs on the server.
|
| Option: | ClientJobBMITimeoutSecs |
| Type: | Integer |
| Contexts: | Defaults |
| Default Value: | 300 |
| Description: | Specifies the timeout value in seconds for BMI jobs on the client.
|
| Option: | ClientJobFlowTimeoutSecs |
| Type: | Integer |
| Contexts: | Defaults |
| Default Value: | 300 |
| Description: | Specifies the timeout value in seconds for FLOW jobs on the client.
|
| Option: | ClientRetryLimit |
| Type: | Integer |
| Contexts: | Defaults |
| Default Value: | 5 |
| Description: | Specifies the number of retry attempts for operations (when possible)
|
| Option: | ClientRetryDelayMilliSecs |
| Type: | Integer |
| Contexts: | Defaults |
| Default Value: | 2000 |
| Description: | Specifies the delay in milliseconds to wait between retries.
|
| Option: | PerfUpdateInterval |
| Type: | Integer |
| Contexts: | Defaults |
| Default Value: | 1000 |
| Description: | This specifies the frequency (in milliseconds)
that performance monitor should be updated
when the pvfs server is running in admin mode. Can be set in either Default or ServerOptions contexts.
|
| Option: | BMIModules |
| Type: | List |
| Contexts: | Defaults |
| Default Value: | None |
| Description: | List the BMI modules to load when the server is started. At present,
only tcp, infiniband, and myrinet are valid BMI modules.
The format of the list is a comma separated list of one of: bmi_tcp
bmi_ib
bmi_gm For example: BMIModules bmi_tcp,bmi_ib Note that only the bmi modules compiled into pvfs should be
specified in this list. The BMIModules option can be specified
in either the Defaults or ServerOptions contexts.
|
| Option: | FlowModules |
| Type: | List |
| Contexts: | Defaults |
| Default Value: | flowproto_multiqueue |
| Description: | List the flow modules to load when the server is started. The modules
available for loading currently are: flowproto_multiqueue - A flow module that handles all the possible flows,
bmi->trove, trove->bmi, mem->bmi, bmi->mem. At present, this is the
default and only available flow for production use. flowproto_bmi_cache - A flow module that enables the use of the NCAC
(network-centric adaptive cache) in the pvfs server. Since the NCAC
is currently disable and unsupported, this module exists as a proof
of concept only. flowproto_dump_offsets - Used for debugging, this module allows the
developer to see what/when flows are being posted, without making
any actual BMI or TROVE requests. This should only be used if you
know what you're doing. |
| Option: | LogStamp |
| Type: | String |
| Contexts: | Defaults
|
| Default Value: | usec |
| Description: | Specifies the format of the date/timestamp that events will have
in the event log. Possible values are: usec: [%H:%M:%S.%U] datetime: [%m/%d %H:%M] thread: [%H:%M:%S.%U (%lu)] none The format of the option is one of the above values. For example, LogStamp datetime
|
| Option: | FlowBufferSizeBytes |
| Type: | Integer |
| Contexts: | FileSystem |
| Default Value: | 262144 |
| Description: | --- end default options for all servers |
| Option: | FlowBuffersPerFlow |
| Type: | Integer |
| Contexts: | FileSystem |
| Default Value: | 8 |
| Description: | number of buffers to use for bulk data transfers |
| Option: | RootSquash |
| Type: | List |
| Contexts: | |
| Default Value: | |
| Description: |
File-system export options Define options that will influence the way a file-system gets exported
to the rest of the world.
|
| Option: | RootSquashExceptions |
| Type: | List |
| Contexts: | |
| Default Value: | |
| Description: | RootSquashExceptions option specifies exceoptions to the RootSquash
list. This is an optional parameter that needs to be specified as
part of the ExportOptions context and is a list of BMI URL
specification of client addresses for which RootSquash
has to be enforced.
RootSquash tcp://192.168.2.0@24 tcp://10.0.0.* tcp://192.168.* ...
|
| Option: | ReadOnly |
| Type: | List |
| Contexts: | |
| Default Value: | |
| Description: | ReadOnly option specifies whether the exported file-system needs to
disallow write accesses from clients or anything that modifies the
state of the file-system.
This is an optional parameter that needs to be specified as part of
the ExportOptions context and is a list of BMI URL specification of
client addresses for which ReadOnly has to be enforced.
An example: ReadOnly tcp://192.168.2.0@24 tcp://10.0.0.* tcp://192.168.* ...
|
| Option: | AllSquash |
| Type: | List |
| Contexts: | |
| Default Value: | |
| Description: | AllSquash option specifies whether the exported file-system needs to
squash all accesses to the file-system to a specified uid/gid!
This is an optional parameter that needs to be specified as part of
the ExportOptions context and is a list of BMI URL specification of client
addresses for which AllSquash has to be enforced.
An example: AllSquash tcp://192.168.2.0@24 tcp://10.0.0.* tcp://192.168.* ...
|
| Option: | AnonUID |
| Type: | String |
| Contexts: | |
| Default Value: | 65534 |
| Description: | AnonUID and AnonGID are 2 integers that tell the servers to translate
the requesting clients' uid/gid to the specified ones whenever AllSquash
is specified!
If these are not specified and AllSquash is specified then the uid used
will be that of nobody and gid that of nobody.
An example: AnonUID 3454
AnonGID 3454
|
| Option: | AnonGID |
| Type: | String |
| Contexts: | |
| Default Value: | 65534 |
| Description: | |
| Option: | HandleRecycleTimeoutSecs |
| Type: | Integer |
| Contexts: | StorageHints |
| Default Value: | 360 |
| Description: | The TROVE storage layer has a management component that deals with
allocating handle values for new metafiles and datafiles. The underlying
trove module can be given a hint to tell it how long to wait before
reusing handle values that have become freed up (only deleting files will
free up a handle). The HandleRecycleTimeoutSecs option specifies
the number of seconds to wait for each filesystem. This is an
optional parameter that can be specified in the StorageHints context.
|
| Option: | AttrCacheKeywords |
| Type: | List |
| Contexts: | StorageHints |
| Default Value: | DATAFILE_HANDLES_KEYSTR |
| Description: | The TROVE layer (server side storage layer)
has an attribute caching component that
caches stored attributes. This is used to improve the performance of
metadata accesses. The AttrCacheKeywords option is a list of the
object types that should get cached in the attribute cache.
The possible values for this option are: dh - (datafile handles) This will cache the array of datafile handles for
each logical file in this filesystem md - (metafile distribution) This will cache (for each logical file)
the file distribution information used to create/manage
the datafiles. de - (directory entries) This will cache the handles of
the directory entries in this filesystem st - (symlink target) This will cache the target path
for the symbolic links in this filesystem The format of this option is a comma-separated list of one or more
of the above values. For example: AttrCacheKeywords dh,md,de,st
|
| Option: | AttrCacheSize |
| Type: | Integer |
| Contexts: | StorageHints |
| Default Value: | 511 |
| Description: | The attribute cache in the TROVE layer mentioned in the documentation
for the AttrCacheKeywords option is managed as a hashtable. The
AttrCacheSize adjusts the number of buckets that this hashtable contains.
This value can be adjusted for better performance. A good hashtable
size should always be a prime number.
|
| Option: | AttrCacheMaxNumElems |
| Type: | Integer |
| Contexts: | StorageHints |
| Default Value: | 1024 |
| Description: | This option specifies the max cache size of the attribute cache
in the TROVE layer mentioned in the documentation
for the AttrCacheKeywords option. This value can be adjusted for
better performance.
|
| Option: | TroveSyncMeta |
| Type: | String |
| Contexts: | StorageHints |
| Default Value: | yes |
| Description: | The TroveSyncMeta option allows users to turn off metadata
synchronization with every metadata write. This can greatly improve
performance. In general, this value should probably be set to yes,
otherwise metadata transaction could be lost in the event of server
failover.
|
| Option: | TroveSyncData |
| Type: | String |
| Contexts: | StorageHints |
| Default Value: | yes |
| Description: | The TroveSyncData option allows users to turn off datafile
synchronization with every write operation. This can greatly improve
performance, but may cause lost data in the event of server failover.
|
| Option: | DBCacheSizeBytes |
| Type: | Integer |
| Contexts: | StorageHints |
| Default Value: | 0 |
| Description: | |
| Option: | DBCacheType |
| Type: | String |
| Contexts: | StorageHints |
| Default Value: | sys |
| Description: | cache type for berkeley db environment. "sys" and "mmap" are valid
value for this option
|
| Option: | Param |
| Type: | String |
| Contexts: | Distribution |
| Default Value: | None |
| Description: | This option specifies a parameter name to be passed to the
distribution to be used. This option should be immediately
followed by a Value option.
|
| Option: | Value |
| Type: | Integer |
| Contexts: | Distribution |
| Default Value: | None |
| Description: | This option specifies the value of the parameter who's name
was specified in the previous option.
|
| Option: | DefaultNumDFiles |
| Type: | Integer |
| Contexts: | FileSystem |
| Default Value: | 0 |
| Description: | This option specifies the default number of datafiles to use
when a new file is created. The value is passed to the distribution
and it determines whether to use that value or not.
|
| Option: | ImmediateCompletion |
| Type: | String |
| Contexts: | StorageHints |
| Default Value: | no |
| Description: | |
| Option: | CoalescingHighWatermark |
| Type: | String |
| Contexts: | StorageHints |
| Default Value: | 8 |
| Description: | |
| Option: | CoalescingLowWatermark |
| Type: | Integer |
| Contexts: | StorageHints |
| Default Value: | 1 |
| Description: | |
| Option: | TroveMethod |
| Type: | String |
| Contexts: | Defaults StorageHints |
| Default Value: | dbpf |
| Description: | This option specifies the method used for trove. Currently the
dbpf method is the default. Other methods include 'alt-aio'. Note that this option can be specified in either the
Defaults context of the main fs.conf, or in a filesystem specific
StorageHints
context, but the semantics of TroveMethod in the
Defaults
context is different from other options. The TroveMethod in the
Defaults context only specifies which
method is used at
server initialization. It does not specify the default TroveMethod
for all the filesystems the server supports. To set the TroveMethod
for a filesystem, the TroveMethod must be placed in the
StorageHints context for that filesystem.
|
| Option: | SecretKey |
| Type: | String |
| Contexts: | FileSystem |
| Default Value: | None |
| Description: | Specifies the file system's key for use in HMAC-based digests of
client operations.
|
Context List
This is the list of possible Contexts that can be used in the config
files in this version of PVFS2.
| Context: | Defaults |
| Parent Context: | Global |
| Description: | Options specified within the Defaults context are used as
default values over all the pvfs2 server specific config files.
|
| Context: | StorageHints |
| Parent Context: | FileSystem |
| Description: | This groups
options specific to a filesystem and related to the behavior of the
storage system. Mostly these options are passed directly to the
TROVE storage module which may or may not support them. The
DBPF module (currently the only TROVE module available) supports
all of them.
|
| Context: | Global |
| Parent Context: | None |
| Description: | Global Context |
| Context: | Security |
| Parent Context: | Global |
| Description: | Options specified within the Security context are used to dictate
whether the pvfs2
servers will accept or handle file-system requests.
This section is optional and does not need to be specified.
|
| Context: | ExportOptions |
| Parent Context: | FileSystem |
| Description: | Specifies the beginning of a ExportOptions context.
This groups options specific to a filesystem and related to the behavior
of how it gets exported to various clients. Most of these options
will affect things like what uids get translated to and so on..
|
| Context: | ServerOptions |
| Parent Context: | Global |
| Description: | This groups the Server specific options. The ServerOptions context should be defined after the Alias mappings
have been defined. The reason is that the ServerOptions context is
defined in terms of the aliases defined in that context. Default options applicable to all servers can be overridden on
a per-server basis in the ServerOptions context.
To illustrate:
Suppose the Option name is X, its default value is Y,
and one wishes to override the option for a server to Y'.
..
X Y
..
Server {server alias}
..
X Y'
..
The ServerOptions context REQUIRES the Server option specify
the server alias, which sets the remaining options specified
in the context for that server.
|
| Context: | DataHandleRanges |
| Parent Context: | FileSystem |
| Description: | This context groups together the Range options that define valid values
for the data handles on a per-host basis for this filesystem. A DataHandleRanges context is required to be present in a
Filesystem context.
|
| Context: | |
| Parent Context: | None |
| Description: | |
| Context: | MetaHandleRanges |
| Parent Context: | FileSystem |
| Description: | This context groups together the Range options that define valid values
for meta handles on a per-host basis for this filesystem. The MetaHandleRanges context is required to be present in a
Filesystem context.
|
| Context: | |
| Parent Context: | None |
| Description: | |
| Context: | Distribution |
| Parent Context: | FileSystem |
| Description: | Provides a context for defining the filesystem's default
distribution to use and the parameters to be set for that distribution. Valid options within the Distribution context are Name, Param, and Value. This context is an optional context within the Filesystem context. If
not specified, the filesystem defaults to the simple-stripe distribution.
|
| Context: | Aliases |
| Parent Context: | Global |
| Description: | This groups the Alias mapping options. The Aliases context should be defined before any FileSystem contexts
are defined, as options in the FileSystem context usually need to
reference the aliases defined in this context.
|
| Context: | FileSystem |
| Parent Context: | Global |
| Description: | This groups options specific to a filesystem. A pvfs2 server may manage
more than one filesystem, so a config file may have more than
one Filesystem context, each defining the parameters of a different
Filesystem.
|
|
Quick Links
-
FAQ
- Frequently Asked Questions.
- Fisheye
- Browse the source code.
- Tinderbox
- Continuous Integration Testing. View the latest test runs and results.
|
|