GNU tar archive tool reference by example

Here:

The file gzipped-folder.tar.gz will contain all the files (including dot files) and subfolders of the current folder.

There are few ways to do it. The older versions of tar do not accept compression level for the gzip, so we have to hint to the gzip in other way.

Same as the above, but use bzip2 compression instead of the gzip. In the past the bzip2 compression produced smaller size archives compared to the gzip, but today they perform pretty much the same.

Here:

The file gzipped-folder.tar.bz2 will contain all the files (including dot files) and subfolders of the current folder.

E.g. create an archive named tared-folder.tar to include all files/subfolders of the current folder except the file named cookbook.gzip and subfolder and its contents named .git:

v is for verbose output during the operation.

Use -t option before the f:

When running tar as scheduled/cron-ed job, it is benefitial to include date of the archive creation in the name.

E.g.: create a tar archive named backup-<current date>.tar from files in the current folder ending in *.md:

Result:

The file(s) will be appended at the end of the archive, just so you know.

E.g. let’s append to the existing backup-13-07-2021.tar archive the file named missfont.log:

An old trick to compensate for various deficiencies of cp and mv.

We just pipe the tar output to the OpenSSL, provided it is already installed. The password is given interactively in the CLI, so this is not very secure way to do so.

E.g. tar the current folder into tar archive and the encrypt it:

Now, decrypt it:

We may specify a specific filename to extrtact or use shell globbing patterns for file name matching.

E.g.: extract only file named README.md from the archive tar cookbooks.tar.bz2:

E.g.: extract all Markdown files from the archive:

Task: add to tar archive and compress contents of the directory ASM on the remote server 19.23.55.158 and download it to the local host as file ASM.tar.gz

Result:

Here:

By default tar will add files/directories to the archive along with their owner user/group. The only reliable way to prevent this is to replace actual data with fake user/group when adding to the archive.

E.g. Add file README.md to the archive, but change the owner’s username/group to the fictitious Doe with numeric id of 1002. If we supply just username/group name, then depending on version/implementation, the tar may change them as asked but leave the real numeric IDs. To force tar not to do it, specify both - alphanumeric name and numeric ID or add beyond numeric IDs the option --numeric-owner, which forces tar to keep only numeric IDs.

Verify:

Not really possible. There is --delete option that seemingly does this, but under the surface this option just combines extracting the whole archive to the temporary directory, deleting the file(s) in question, and creating the archive again from scratch into one command.

The short answer - you can’t. The extended answer - you can’t archive in parallel to the same archive (it was never the goal of tar, which originally wrote archives to the physical tapes that could not be accessed in parallel), but you have options (if you need at all) to parallelize compression of the archive. The options for parallel execution depend on the compressing utility used. There are xz, 7zip, and pigz tools which can compress an archive in parallel, given the correct options. But they cannot decompress in parallel way though, only to compress.

In situation where you are presented with a bunch of files with random names, finding which ones are proper tar archive can be done in few ways. The idea behind all of them is to look for the tar’s magic number inside the file. On systems with file utility installed, it is really easy:

As you can see, it found tar archives without any extension permstar and permstar2.

When the file tool is not available (highly unprobable), we can go more old school way looking at the magic number:

Here:

Use -h switch to tell tar to dereference symlinks and add to archive objects (directories/files) that those symlinks point to.

This will dereference all symlinks found in the current directory.

Tar itself does not have option to search by timestamps, but find does.

Here:

Tar itself does not calculate/save checksum in the archive it creates. The rudimentary "integrity" check can be done with -t switch, which produces an error and exits if the archive is severely damaged - cannot be read, headers are mangled and such. The change in the contents of a file this -t check will NOT notice. When gzip-ing tar archive, though, the CRC checksum is autosaved, but of the final tar archive, not individual files inside this archive. This way, if there is a CRC checksum mismatch on unzipping tar archive, the gzip will issue an error on the standard output.

So, to try and read the archive, verifying that it is readable:

Follow me on https://www.linkedin.com/in/yurislobodyanyuk/ or https://github.com/yuriskinfo not to miss what I publish on Linkedin, Github, blog, and more.