LTFS Data Management
Overview
With LTFS Data Management it is possible to add tape within a storage hierarchy to disk. This is possible for Linux file system types that are compliant to the POSIX standard. If file is migrated from disk to tape data is removed from the disk file system and stored on tape. By using a Fuse overlay file system on top of the original disk file system, the original files name space and the original file sizes are exposed to applications. Accessing the file using read, write, or truncate system calls will trigger a recall to transfer back the data to disk. The corresponding system calls are blocked until the whole data is transferred.
Prerequisites
Because of other dependencies LTFS Data Management only is possible to operate at the following operating system:
- Red Hat Enterprise Linux Server release 7.X
To build and run the code there are dependencies on other software. The following software needs to be installed prior to build and run the LTFS Data Management code:
- non RHEL 7 software:
- IBM Tape Device Driver at: Fix Central. Follow the documentation of this software for installation.
- Spectrum Archive Library Edition version 2.4.0.X available at: Fix Central. This software will be installed in two steps. Follow the installation instructions specified within its documentation.
- RHEL 7 packages:
- protobuf
- protobuf-compiler
- protobuf-devel
- sqlite
- sqlite-devel
- fuse
- fuse-devel
- libuuid
- libuuid-devel
- libmount
- libmount-devel
- libxml2
- libxml2-devel
- boost-system
- boost-thread
- boost-system
- boost-devel
Build
To compile the code the GCC C++ compiler is required version 4.8.5 or higher that supports the C++11 standard. If all prerequisite software is installed and the code is extracted to a directory like the following:
<path to code>/LTFS-Data-Management
change to that directory and run the make command. The resulting executables are located within the
<path to code>/LTFS-Data-Management/bin
directory.
Design documentation
The documentation is include within the source code in a Doxygen format. It has to be compiled after the code has been built. The doxygen RHEL 7 software package has to be installed prior to compile the documentation. The compilation happens by executing the doxygen command within the
<path to code>/LTFS-Data-Management
directory. Error messages can be ignored. To read the documentation use a browser and access the
<path to code>/LTFS-Data-Management/html/index.html
file.
A copy of the documentation is available. It might not correspond to the current code level:
LTFS Data Management documentation
Setting up the environment
To run the code the following shell variables need to be extended by the following:
| variable | extension |
|---|---|
| PATH | <path to code>/LTFS-Data-Management/bin |
| LD_LIBRARY_PATH | /usr/local/lib:/opt/IBM/ltfs/lib64:<path to code>/LTFS-Data-Management/bin |
Starting the software
Prior to starting LTFS Data Management, Spectrum Archive Library Edition needs to be started. A directory needs to be created as a mount point for LTFS. The name of this directory can be freely chosen. If e.g. the directory name is "/ltfs" and the corresponding IBM changer device name is "/dev/IBMchanger0" (look at the documentation for the IBM Tape Device drive - a prerequisite to the Spectrum Archive software - for more information), the Spectrum Archive software needs to be started with the following options:
/opt/IBM/ltfs/bin/ltfs /ltfs -o changer_devname=/dev/IBMchanger0 -o sync_type=unmount -o manual_mount_mode
After Spectrum Archive Library Edition is running LTFS Data Management can be started in the following way:
ltfsdm start
Contact information
Martin Petermann (MAP@zurich.ibm.com)