OneinStack: one shell script, a version catalogue, and a cron entry with a stray ampersand
OneinStack - A PHP/JAVA Deployment Tool
At a glance
- What is it?
- This is a shell installer for the usual Linux web stack, and it is thorough in a way that is easy to under-read. It names eighteen database builds, fifteen PHP versions, five web servers and six JDKs, supports distributions going back to releases that have been out of support for years, and keeps its whole version matrix in a text file at the root. The documentation has two small defects that anyone following it will hit.
- Who is it for?
- Adopt OneinStack when you want a working stack in one command on a distribution it names and you want to choose exact component versions, because the version matrix is explicit and the installation and data directories are configurable in one file before anything runs.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 3 days ago.
- What is it written in?
- Mainly Shell, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- oneinstack/oneinstack2,462 ★ · +5
| Repository | 2026-09-30 | 2026-11-15 |
|---|---|---|
| oneinstack/oneinstack | 2.5k | 2.5k |
5 Stars in the last 46 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
The whole version matrix is a text file at the root
The most important file in the repository is not a script. It is a version catalogue at the root, next to the installer, and it is what lets a shell script know what to download.
The component lists are long. Databases number eighteen: six MySQL builds including two labelled as long-term support, five MariaDB, four Percona, one PostgreSQL, plus ClickHouse and MongoDB. PHP runs from 5.3 to 8.5 in fifteen steps. Tomcat goes from 7 to 11. Web servers include Nginx with native HTTP/3 and QUIC on a stated OpenSSL version, Tengine, OpenResty, Caddy, Apache and a Lua-based web application firewall.
The Java list is the one to read twice: OpenJDK 8, 11, 17 and 18. It stops at 18, so the three versions released after it are not offered, and 17 is the one that would be current for a long-running service.
Source builds are the default approach, with the stated rationale that the latest source is the most stable, fetched from the official host or a mirror. That is a deliberate choice against distribution packages, and it means the script compiles what it installs rather than accepting what your distribution ships.
The distribution list reaches back past end of life
The support statement is one long sentence and it covers a lot of ground.
It lists 64-bit Red Hat Enterprise Linux 7, 8 and 9, with CentOS, Red Hat, AlmaLinux, Rocky and Anolis named as family members, then Debian 9 through 13, Ubuntu 16 through 24 in four-year steps, TencentOS in two ranges, openEuler, Kylin, and Fedora from 27 onwards.
Read as a coverage statement that is impressive. Read as a testing statement it says something different, because several of those entries have been unsupported for years: Ubuntu 16 and Debian 9 are both long past their end of life, and the 64-bit restriction also means nothing here runs on a 32-bit host.
The stack itself is equally broad. The five deployment profiles in the name are combinations of a web server from a set of five, a database from the eighteen above, PHP, and an optional Java runtime, with an in-memory store chosen at install time between Valkey, described as the open-source alternative, Redis, or Memcached.
So the choice you make at install time is not one binary but a matrix, and the script's job is to resolve it.
The cron entry in the backup section ends with an ampersand
The backup section is the most practically useful part of the documentation and also the most flawed.
The commands are laid out with inline annotations:
~/oneinstack/backup_setup.sh // Backup parameters
~/oneinstack/backup.sh // Perform the backup immediately
crontab -l // Can be added to scheduled tasks, such as automatic backups every day 1:00
0 1 * * * cd ~/oneinstack/backup.sh > /dev/null 2>&1 &Two things are wrong with that block. The annotations use a double slash, which is not a shell comment, so a reader who copies the lines as written gets a command with arguments. And the schedule line ends with an ampersand, which is a habit from an interactive shell where you want to return to a prompt; a cron entry is already run in its own shell, so the ampersand adds nothing and makes the entry harder to reason about.
The intent is clear and reasonable: run the backup script daily at one in the morning, discard its output, and inspect the existing schedule first. The script pair makes sense too, with one for configuration and one for execution.
What the section does not mention is which destinations are available. The feature list names them: local, remote between servers over rsync, and five object storage providers plus two consumer drive services.
Two sections in the README are titled Installation
The document has a duplicated heading, and the second one is not an installation section at all.
The first installation section is the real one: it installs the two dependencies, downloads a tarball, extracts it, and runs the installer. The second, near the end of the document, is titled the same way and contains a line about feedback and questions, a Telegram group link and the project site.
That is a copy-paste artefact in a document that is otherwise generated by hand, and its effect is navigational rather than technical: a table of contents built from those headings would show two identically named entries, and a reader looking for where to get help has to read a section that promises installation steps.
It is worth noting because this document is the entire user-facing surface of the project. There is no separate documentation site linked from it, and the repository has two READMEs, one in English and one in Chinese, so any structural mistake is duplicated in both.
options.conf is the file to edit before install.sh runs
The install sequence is short and has one deliberate step.
You install the two dependencies the script needs, which differ by distribution family, then download a tarball from the project's mirror host, extract it, and change into the directory. If your connection drops mid-install, the documentation says you can reattach to the install window with a screen command.
The step before running the installer is the interesting one. If you need to change the installation directory, the data storage location or where Nginx writes its logs, you edit a configuration file named options.conf, and then run the installer. So the script's destinations are decided before anything is installed rather than through flags at install time.
The screen instructions have a small inconsistency. The prose explains that one command reattaches to a detached install window, and the code block next to it uses the flag that names a new session rather than the one that reattaches. Both forms appear within a few lines of each other and only one is explained, which is the same class of problem as the duplicated heading: a hand-maintained document with small breaks in it.
Afterwards, adding a second PHP version is a flag on the same installer, with a two-digit version number in the old scheme.
The script list is the real interface
The top level of the repository is a set of task scripts, and the README is organised around them rather than around concepts.
There is the installer, an add-ons script for extensions, a virtual host script that takes a delete flag to remove one rather than add it, a separate FTP virtual host script, a backup pair split between configuration and execution, an upgrade script and an uninstall script. There is also a root database password reset script, which is the one with the most operational weight and the least documentation.
The rest of the tree is the machinery those scripts use: an include directory, a source directory, a tools directory, an init directory for service units, and a configuration directory. So the structure follows the installer rather than a library layout, which is what you would expect from a set of provisioning scripts and what makes them hard to reuse one at a time.
Service management is documented per component with systemd commands. Nginx, MySQL, PHP, Apache, Memcached and Redis get a command that includes a reload verb. PostgreSQL, MongoDB, Tomcat, Pure-FTPD, Valkey and ClickHouse do not, because those units do not support it. The distinction is real and the documentation happens to get it right in each case, which is more than the two earlier sections manage.
Editorial conclusion
Adopt OneinStack when you want a working stack in one command on a distribution it names and you want to choose exact component versions, because the version matrix is explicit and the installation and data directories are configurable in one file before anything runs. Do not adopt it expecting a version-agnostic installer, since the release cadence has been irregular across three years and the supported distribution list includes releases past their end of life, which tells you what is tested rather than what is maintained. Before you run it, read the options file, and if you set up the backup cron entry, take the trailing ampersand off the line the README gives you.
Frequently asked questions
What does OneinStack install?
A Linux stack combining a web server from Nginx, Tengine, OpenResty, Caddy or Apache with a database, PHP, an optional Java runtime and a choice of in-memory store. The profiles are named for those combinations, and the installation and data directories are set in a configuration file you edit before running the installer.
Which component versions does OneinStack offer?
Eighteen database builds across MySQL, MariaDB, Percona, PostgreSQL, ClickHouse and MongoDB, fifteen PHP versions from 5.3 to 8.5, Tomcat 7 through 11, and four JDK builds. The catalogue itself lives in a versions file at the repository root, and installs compile from source rather than taking distribution packages.
Which Linux distributions does OneinStack support?
Sixty-four-bit Red Hat Enterprise Linux 7 through 9 and its family members, Debian 9 through 13, Ubuntu 16 through 24, TencentOS, openEuler, Kylin and Fedora 27 or later. Several of those releases are past their end of support, so the list is better read as coverage than as a maintenance statement.
How do I set up automatic backups with OneinStack?
Configure the backup with the setup script and run the backup script to perform one immediately, then add the execution script to cron. The README's example schedule line ends with a trailing ampersand that serves no purpose in cron, and its inline annotations use a double slash rather than a shell comment.
How do I upgrade or remove a OneinStack installation?
There is a dedicated upgrade script and a dedicated uninstall script, both documented as one command each from the installation directory. Adding a second PHP version uses the installer itself with a version flag in the project's two-digit numbering scheme.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/oneinstack-oneinstack)Star history chart for your README
[](https://hysenlabs.com/projects/oneinstack-oneinstack)Was this useful?