Administration Console Online Help

Previous Next Open TOC in new window
Content starts here

File Store: Configuration

Configuration Options     Related Tasks     Related Topics

Use this page to configure a disk-based file store for storing subsystem data, such as persistent JMS messages or Store-and-Forward messages.

Configuration Options

Name Description
Name

The name of this file store. This name must be unique within the WebLogic Server instance or its cluster.

MBean Attribute:
WebLogicMBean.Name

Changes take effect after you redeploy the module or restart the server.

Synchronous Write Policy

The disk write policy that determines how the file store writes data to disk.

This policy also affects the JMS file store's performance, scalability, and reliability. Oracle recommends Direct-Write-With-Cache which tends to have the highest performance. The default value is Direct-Write. The valid policy options are:

  • Direct-Write

    Direct I/O is supported on all platforms. When available, file stores in direct I/O mode automatically load the native I/O wlfileio driver. This option tends to out-perform Cache-Flush and tend to be slower than Direct-Write-With-Cache. This mode does not require a native store wlfileio driver, but performs faster when they are available.

  • Direct-Write-With-Cache

    Store records are written synchronously to primary files in the directory specified by the Directory attribute and asynchronously to a corresponding cache file in the Cache Directory. The Cache Directory provides information about disk space, locking, security, and performance implications. This mode requires a native store wlfileiocode driver. If the native driver cannot be loaded, then the write mode automatically switches to Direct-Write. See Cache Directory.

  • Cache-Flush

    Transactions cannot complete until all of their writes have been flushed down to disk. This policy is reliable and scales well as the number of simultaneous users increases.Transactionally safe but tends to be a lower performer than direct-write policies.

  • Disabled

    Transactions are complete as soon as their writes are cached in memory, instead of waiting for the writes to successfully reach the disk. This is the fastest policy because write requests do not block waiting to be synchronized to disk, but, unlike other policies, is not transactionally safe in the event of operating system or hardware failures. Such failures can lead to duplicate or lost data/messages. This option does not require native store wlfileio drivers, but may run faster when they are available. Some non-WebLogic JMS vendors default to a policy that is equivalent to Disabled.

Notes:

  • When available, file stores load WebLogic wlfileio native drivers, which can improve performance. These drivers are included with Windows, Solaris, Linux, HP-UX, and AIX WebLogic installations.

  • Certain older versions of Microsoft Windows may incorrectly report storage device synchronous write completion if the Windows default Write Cache Enabled setting is used. This violates the transactional semantics of transactional products (not specific to Oracle), including file stores configured with a Direct-Write (default) or Direct-Write-With-Cache policy, as a system crash or power failure can lead to a loss or a duplication of records/messages. One of the visible symptoms is that this problem may manifest itself in high persistent message/transaction throughput exceeding the physical capabilities of your storage device. You can address the problem by applying a Microsoft supplied patch, disabling the Windows Write Cache Enabled setting, or by using a power-protected storage device. See http://support.microsoft.com/kb/281672 and http://support.microsoft.com/kb/332023.

  • NFS storage note: On some operating systems, native driver memory-mapping is incompatible with NFS when files are locked. Stores with synchronous write policies Direct-Write-With-Cache or Disabled, and WebLogic JMS paging stores enhance performance by using the native wlfileio driver to perform memory-map operating system calls. When a store detects an incompatibility between NFS, file locking, and memory mapping, it automatically downgrades to conventional read/write system calls instead of memory mapping. For best performance, Oracle recommends investigating alternative NFS client drivers, configuring a non-NFS storage location, or in controlled environments and at your own risk, disabling the file locks (See Enable File Locking). For more information, see "Tuning the WebLogic Persistent Store" in Performance and Tuning for Oracle WebLogic Server.

MBean Attribute:
GenericFileStoreMBean.SynchronousWritePolicy

Target

The server instances or migratable targets defined in the current domain that are candidates for hosting the file store.

In a clustered environment, a recommended best practice is to target a custom file store to the same migratable target as the migratable JMS service, so that a member server will not be a single point of failure. A file store can also be configured to automatically migrate from an unhealthy server instance to a healthy server instance with the help of the server health monitoring services.

MBean Attribute:
PersistentStoreMBean.Targets

Directory

The path name to the file system directory where the file store maintains its data files.

  • When targeting a file store to a migratable target, the store directory must be accessible from all candidate server members in the migratable target.

  • For highest availability, use either a SAN (Storage Area Network) or other reliable shared storage.

  • Use of NFS mounts is discouraged, but supported. Most NFS mounts are not transactionally safe by default, and, to ensure transactional correctness, need to be configured using your NFS vendor documentation in order to honor synchronous write requests.

  • For SynchronousWritePolicy of Direct-Write-With-Cache, see Cache Directory.

  • Additional O/S tuning may be required if the directory is hosted by Microsoft Windows, see Synchronous Write Policy for details.

MBean Attribute:
FileStoreMBean.Directory

Changes take effect after you redeploy the module or restart the server.

Logical Name

The name used by subsystems to refer to different stores on different servers using the same name.

For example, an EJB that uses the timer service may refer to its store using the logical name, and this name may be valid on multiple servers in the same cluster, even if each server has a store with a different physical name.

Multiple stores in the same domain or the same cluster may share the same logical name. However, a given logical name may not be assigned to more than one store on the same server.

MBean Attribute:
FileStoreMBean.LogicalName

Cache Directory

The location of the cache directory for Direct-Write-With-Cache, ignored for other policies.

When Direct-Write-With-Cache is specified as the SynchronousWritePolicy, cache files are created in addition to primary files (see Directory for the location of primary files). If a cache directory location is specified, the cache file path is CacheDirectory/WLStoreCache/StoreNameFileNum.DAT.cache. When specified, Oracle recommends using absolute paths, but if the directory location is a relative path, then CacheDirectory is created relative to the WebLogic Server instance's home directory. If "" or Null is specified, the Cache Directory is located in the current operating system temp directory as determined by the java.io.tmpdir Java System property (JDK's default: /tmp on UNIX, %TEMP% on Windows) and is TempDirectory/WLStoreCache/DomainName/unique-id/ StoreNameFileNum.DAT.cache. The value of java.io.tmpdir varies between operating systems and configurations, and can be overridden by passing -Djava.io.tmpdir=My_path on the JVM command line.

Considerations:

  • Security: Some users may want to set specific directory permissions to limit access to the cache directory, especially if there are custom configured user access limitations on the primary directory. For a complete guide to WebLogic security, see "Securing a Production Environment for Oracle WebLogic Server."

  • Additional Disk Space Usage: Cache files consume the same amount of disk space as the primary store files that they mirror. See Directory for the location of primary store files.

  • Performance: For the best performance, a cache directory should be located in local storage instead of NAS/SAN (remote) storage, preferably in the operating system's temp directory. Relative paths should be avoided, as relative paths are located based on the domain installation, which is typically on remote storage. It is safe to delete a cache directory while the store is not running, but this may slow down the next store boot.

  • Preventing Corruption and File Locking: Two same named stores must not be configured to share the same primary or cache directory. There are store file locking checks that are designed to detect such conflicts and prevent corruption by failing the store boot, but it is not recommended to depend on the file locking feature for correctness. See Enable File Locking.

  • Boot Recovery: Cache files are reused to speed up the File Store boot and recovery process, but only if the store's host WebLogic Server instance has been shut down cleanly prior to the current boot. For example, cache files are not re-used and are instead fully recreated: after a kill -9, after an OS or JVM crash, or after an off-line change to the primary files, such as a store admin compaction. When cache files are recreated, a Warning log message 280102 is generated.

  • Fail-Over/Migration Recovery: A file store safely recovers its data without its cache directory. Therefore, a cache directory does not need to be copied or otherwise made accessible after a fail-over or migration, and similarly does not need to be placed in NAS/SAN storage. A Warning log message 280102, which is generated to indicate the need to recreate the cache on the new host system, can be ignored.

  • Cache File Cleanup: To prevent unused cache files from consuming disk space, test and developer environments should periodically delete cache files.



MBean Attribute:
FileStoreMBean.CacheDirectory

Minimum Window Buffer Size

The minimum amount of data, in bytes and rounded down to the nearest power of 2, mapped into the JVM's address space per primary store file. Applies to synchronous write policies Direct-Write-With-Cache and Disabled, but only when a native wlfileio library is loaded. See Maximum Window Buffer Size.

MBean Attribute:
FileStoreMBean.MinWindowBufferSize

Minimum value: -1

Maximum value: 1073741824

Maximum Window Buffer Size

The maximum amount of data, in bytes and rounded down to the nearest power of 2, mapped into the JVM's address space per primary store file. Applies to synchronous write policies Direct-Write-With-Cache and Disabled but only when the native wlfileio library is loaded.

A window buffer does not consume Java heap memory, but does consume off-heap (native) memory. If the store is unable to allocate the requested buffer size, it allocates smaller and smaller buffers until it reaches MinWindowBufferSize, and then fails if cannot honor MinWindowBufferSize.

Oracle recommends setting the max window buffer size to more than double the size of the largest write (multiple concurrently updated records may be combined into a single write), and greater than or equal to the file size, unless there are other constraints. 32-bit JVMs may impose a total limit of between 2 and 4GB for combined Java heap plus off-heap (native) memory usage.

MBean Attribute:
FileStoreMBean.MaxWindowBufferSize

Minimum value: -1

Maximum value: 1073741824

IO Buffer Size

The I/O buffer size, in bytes, automatically rounded down to the nearest power of 2.

  • For the Direct-Write-With-Cache policy when a native wlfileio driver is available, IOBufferSize describes the maximum portion of a cache view that is passed to a system call. This portion does not consume off-heap (native) or Java heap memory.

  • For the Direct-Write and Cache-Flush policies, IOBufferSize is the size of a per store buffer which consumes off-heap (native) memory, where one buffer is allocated during run-time, but multiple buffers may be temporarily created during boot recovery.

  • When a native wlfileio driver is not available, the setting applies to off-heap (native) memory for all policies (including Disabled).

  • For the best runtime performance, Oracle recommends setting IOBufferSize so that it is larger than the largest write (multiple concurrent store requests may be combined into a single write).

  • For the best boot recovery time performance of large stores, Oracle recommends setting IOBufferSize to at least 2 megabytes.

  • See AllocatedIOBufferBytes to find out the actual allocated off-heap (native) memory amount. It is a multiple of IOBufferSize for the Direct-Write and Cache-Flush policies, or zero.

  • See AllocatedIOBufferBytes.



MBean Attribute:
FileStoreMBean.IoBufferSize

Minimum value: -1

Maximum value: 67108864

Maximum File Size

The maximum file size, in bytes.

  • The MaxFileSize value affects the number of files needed to accommodate a store of a particular size (number of files = store size/MaxFileSize rounded up).

  • A file store automatically reuses space freed by deleted records and automatically expands individual files up to MaxFileSize if there is not enough space for a new record. If there is no space left in exiting files for a new record, a store creates an additional file.

  • A small number of larger files is normally preferred over a large number of smaller files as each file allocates Window Buffer and file handles.

  • If MaxFileSize is larger than 2^24 * BlockSize, then MaxFileSize is ignored, and the value becomes 2^24 * BlockSize. The default BlockSize is 512, and 2^24 * 512 is 8 GB.

  • See Initial Size.



MBean Attribute:
FileStoreMBean.MaxFileSize

Minimum value: 1048576

Enable File Locking

Determines whether OS file locking is used.

When file locking protection is enabled, a store boot fails if another store instance already has opened the store files. Do not disable this setting unless you have procedures in place to prevent multiple store instances from opening the same file. File locking is not required but helps prevent corruption in the event that two same-named file store instances attempt to operate in the same directories. This setting applies to both primary and cache files.

MBean Attribute:
FileStoreMBean.FileLockingEnabled

Block Size

The smallest addressable block, in bytes, of a file. When a native wlfileio driver is available and the block size has not been configured by the user, the store selects the minimum OS specific value for unbuffered (direct) I/O, if it is within the range [512, 8192].

A file store's block size does not change once the file store creates its files. Changes to block size only take effect for new file stores or after the current files have been deleted. See "Tuning the Persistent Store" in Performance and Tuning for Oracle WebLogic Server.

MBean Attribute:
FileStoreMBean.BlockSize

Minimum value: -1

Maximum value: 8192

Initial Size

The initial file size, in bytes.

  • Set InitialSize to pre-allocate file space during a file store boot. If InitialSize exceeds MaxFileSize, a store creates multiple files (number of files = InitialSize/MaxFileSize rounded up).

  • A file store automatically reuses the space from deleted records and automatically expands a file if there is not enough space for a new write request.

  • Use InitialSize to limit or prevent file expansions during runtime, as file expansion introduces temporary latencies that may be noticeable under rare circumstances.

  • Changes to initial size only take effect for new file stores, or after any current files have been deleted prior to restart.

  • See Maximum File Size.



MBean Attribute:
FileStoreMBean.InitialSize

Minimum value: 0

Related Tasks

Related Topics


Back to Top