#server #object-storage #database

bin+lib bulletin-board-server

Object storage for ArrayObject for debugging and data taking purposes

8 releases

new 0.3.2 Dec 9, 2024
0.3.1 Nov 16, 2024
0.2.1 Nov 4, 2024
0.1.2 Oct 18, 2024

#545 in Debugging

Download history 55/week @ 2024-10-07 413/week @ 2024-10-14 15/week @ 2024-10-21 112/week @ 2024-10-28 250/week @ 2024-11-04 141/week @ 2024-11-11 51/week @ 2024-11-18 7/week @ 2024-11-25 3/week @ 2024-12-02

239 downloads per month

Apache-2.0

61KB
1.5K SLoC

Bulletin Board Server

"Buy Me A Coffee"

"Github Sponsors" Crates.io Crates.io License

Object storage for ArrayObject for debugging and data taking purposes.

BulletinBoard is a part of dbgbb project.

Highlights

  • Hybrid backend of memory and file, selected based on the size of the object and the allocated memory.
  • Key is a combination of a title and a tag. Each key contains revisions of ArrayObject.
  • Simple access to data. For example, revision can be omitted. Then, the most recent revision is returned. The tag can also be omitted if no other tags are present.
  • The commands archive and dump make data persistent. (Data does not persist by default.)
  • Docker image of the server is available.
  • Unix sockets can be used with Unix-like operating systems, which makes the communication speed quite fast.
  • GUI application is available. You can download the binary files from Release.

Caution

  • Clients do not check whether the operation is successful or not to improve performance. Check the log of the server for the errors.
  • The data is not encrypted. Please do not send any confidential data over the network.
  • This crate is under development and is subject to change in specification. (Compatibility across BulletinBoard and dbgbb is ensured for the most minor version numbers.)
  • The included tests will access the server and potentially erase existing data.

Usage

There are several options for running the BulletinBoard server.

Cargo

Install and run the server with the specified listen address.

cargo install bulletin-board-server
export BB_LISTEN_ADDR = "0.0.0.0:7578"
bulletin-board-server

Built-in server (GUI)

The server is included in BulletinBoard GUI. Choose Built-in in the settings page and start the server in the start page.

Docker

The docker image is available and you can run the server by

docker run -p 7578:7578 -v /path/to/vol:/data yshojihep/bulletin-board:latest

For details, see DockerHub.

Environment Variables

Variable Default Description
BB_LISTEN_ADDR "127.0.0.1:7578" Listen address of the bulletin board server. If you use a Unix socket, the address should be the path to an uncreated socket.
BB_TMP_DIR "./bb_tmp" Directory for temporary data.
BB_ACV_DIR "./bb_acv" Directory for archives.
BB_TOT_MEM_LIMIT "1GiB" Total memory limit. If the memory exceeds the limit, all the bulletins are saved as files. The size of metadata is not included in the calculation. The actual memry consumption becomes higher than this.
BB_FILE_THRETHOLD "1MiB" Beyond this threthold, the bulletin is saved as a file.
BB_MAX_RESULTS 1024 The maximum number of results returned by viewboard and get_info functions.
BB_LOG_FILE "./bulletin-board.log" Location of the log file.
BB_LOG_LEVEL 3 Log level. The alllowed values are 0: No logging, 1: Error, 2: +Warn, 3: +Notice, 4: +Info, 5: +Debug.
BB_DEBUG Not set If the variable is set, the server logs to stdout.

Command line options

Short Long Description
-d --debug Log to stdout.
-l <LOG_LEVEL> --log-level <LOG_LEVEL> Log level [0: No logging, 1: Error, 2: +Warn, 3: +Notice (default), 4: +Info, 5: +Debug].
-h --help Print help.
-V --version Print version.

Q&A

Why not persistent by default?

Since BulletinBoard was originally designed for debugging purposes, it is assumed that most of the data will be deleted at the end. Persistent options (archive and dump) have been added for a more extensive use such as data taking. The advantages of not making it persistent by default are (i) holding data in memory makes read/write speeds faster, (ii) metadata of the archive becomes smaller and (iii) data can be more easily deleted before archiving.

Why not other object storages or databases?

Especially for debugging, storage may receive large amounts of small data and thus in-memory databases are ideal. However, it may also receive large data like a few hundred MiB, and such data should be stored in files. BulletinBoard uses a hybrid backend of memory and file to solve this problem. Also, the BulletinBoard will not return a response if it is not needed. Thus, it can handle very frequent data flows.

Dependencies

~3–4MB
~71K SLoC