Important tests are being done with FSArchiver to make sure it works as expected. Tests environments have been installed: different Linux distributions (CentOS-5.2, Fedora-8.0, Fedora-11, Ubuntu-9.04) have been installed using different file systems, in order to make sure FSArchiver can compile on various Linux flavors, and that it can save and restore on all these operating systems with no error.
Here is the detailed test procedure that we follow:
During the development, FSArchiver is tested with a Fedora-8 installation. FSArchiver-0.4.7 has been able to successfully save and restore the ext3 root filesystem of a Fedora-8 installation with selinux enabled. This fedora test system has been installed with selinux enabled just because selinux does file labelling on the filesystem using extended attributes, and then it’s a good test to know whether or not these attributes have been preserved by FSArchiver. If you save and restore such a filesystem with a program that does not preserve the extended attributes, selinux will complain when you reboot, and it will make the system unusable. That’s why this is a good test.
An important test is done using rsync. It requires two partitions: the original one, and a spare partition where to restore the archive. It allows to know whether or not there are differences between the original and the restored filesystem. rsync is able to compare both the files contents, and files attributes (timestamps, permissions, owner, extended attributes, acl, …), so that’s a very good test. The following command can be used to know whether or not files are the same (data and attributes) on two file-systems:
rsync -axHAXnP /mnt/part1/ /mnt/part2/
Tests with CentOS-3.9 and CentOS-4.7 have been done to make sure these quite old Linux systems are able to work on a filesystem restored by FSArchiver. When you restore a filesystem using FSArchiver, it may use a very recent version of mkfs to recreate the filesystem. There could be a problem with that. The ext3 filesystem format has been improved over the time, and old Linux system such as CentOS-3.9 running linux-2.4 may not be able to mount the filesystem build with a recent mkfs.ext3. For that reason, FSArchiver preserves the filesystem attributes and features. In other words it excludes the recent ext3 features which may be enabled by default in mkfs.ext3 1.41. That way the filesystem is recreated with its original features, and the old kernel is able to uses this filesystem. Since CentOS is a RHEL clone, it means the test is also valid for that linux flavor. These tests also indicates that in general, the ext3 filesystem attributes are preserved, and then it should work with all other linux distributions.
Here is a command that show the extended attributes of your files:
# getfattr -P -d -R -h -m . /etc/ getfattr: Removing leading '/' from absolute path names # file: etc security.selinux="system_u:object_r:etc_t:s0\000" # file: etc/ntp.conf security.selinux="system_u:object_r:net_conf_t:s0\000" # file: etc/rpm security.selinux="system_u:object_r:etc_t:s0\000" # file: etc/rpm/macros.kde4 security.selinux="system_u:object_r:etc_t:s0\000" # file: etc/rpm/macros.specspo security.selinux="system_u:object_r:etc_t:s0\000"
fsarchiver is using dynamically allocated memory to carry data blocks from one thread to another. There is a risk to have allocated memory which is never freed. For that reason, fsarchiver is tested using valgrind. It tracks all the memory allocations (malloc and free) and prints a message when it detects memory leaks. Here is how to use valgrind that way:
#compile fsarchiver with the debugging option turned on in gcc (gcc -g) #valgrind --tool=memcheck --leak-check=yes --show-reachable=yes --leak-resolution=med --track-origins=yes fsarchiver <fsarchiver-arguments>