Skip to content

giufus/elasticli.rs

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

17 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Elasticli

Logo

Rust

The missing Command Line Interface to interact with Elasticsearch (or yet another one).
Even if elasticsearch APIs are super easy to invoke and very well documented, I always forget the basic commands and methods to do indexing or searches, so I decided to write my own CLI.
I hope you find it useful, it is a work in progress.

Features

Written in Rust 🦀, it uses the following crates:

  • clap for arguments parsing
  • hydroconf for configuration management
  • reqwest for http requests toward elasticsearch

If opportunely configured, the command can run inside an SSH tunnel (with auto close after a number of seconds).

Currently, you can use elasticli to:

  • get info about the target elasticsearch version;
  • create, read, update, delete an index;
  • create, read, update, delete a document;

Changelog

0.1.2: cross platform building
0.1.1: basic authentication
0.1.0: first release

Build & Run

Option 1

  • Build it for your platform with cargo build --release (or just type make), then go to target/release and run the binary elasticli.

Option 2

  • Alternatively, download the pre-built binary for your platform in the releases page. The executable is available for the following targets:

    • x86_64-unknown-linux-gnu -> 64-bit Linux (kernel 3.2+, glibc 2.17+) # CROSS BUILD
    • x86_64-pc-windows-msvc -> 64-bit MSVC (Windows 7+) # BUILT ON WINDOWS ** ***
    • x86_64-apple-darwin -> 64-bit macOS (10.7+, Lion+) # BUILT ON MAC x86_64 **
    • aarch64-unknown-linux-gnu -> ARM64 Linux (kernel 4.1, glibc 2.17+) # CROSS BUILD
    • aarch64-apple-darwin -> ARM64 macOS (11.0+, Big Sur+) # BUILT ON MAC aarch64

**: cross uses docker images to cross compile, unfortunately windows and apple images are not available at the moment.
***: on windows, apart from install rust, you may need to install perl to build elasticli.

You can get the complete list of rust targets with:
rustc --print target-list

Your platform is shown in the host property of the output of:
rustc -vV

Option 3

If you want to experience the cross-compilation, install cross (you may need other dependencies), grab a beer, then run:
make <YOUR-TARGET>
where the target is one of the outputs of rustc --print target-list. Anyway, I have already cross-compiled some of them for you.

Other

  • If you want to run unit tests run cargo test.
  • If you want to build and run directly from sources cargo run -- <your elasticli options and command here>.

Configuration

  • Every command can use the hydroconf features to override default configurations. For example you can override some defaults passing environment name as env var to the command line:
    ENV_FOR_HYDRO=production elasticli info
    or per single prop, specifying it as an env var to the command line:
    HYDRO_ELASTIC__PASSWORD="an even stronger password" elasticli info

Look at the hydroconf doc for major details.

  • You pass the config root directory with -c option before one of the commands (info, index, document). The directory must contain the configuration files (look in the default folder for the latest version):

.secrets.toml environment based sensitive configurations (tipically elastic user and password).

[default]
elastic.username = 'elastic'
elastic.password = 'secure_password'

[production]
elastic.username = 'elastic'
elastic.password = 'changeme'

settings.toml environment based configurations

[default]
# your target elasticsearch host, port, protocol and version
elastic.host = 'otherhost'
elastic.port = 9200
elastic.protocol = 'http'
elastic.version = '8.8.0'

# if enabled, the main command will be executed inside an ssh tunnel. It is the same as running 
###  ssh -i <proxy.key> <proxy.remote_user>@<proxy.host> -f -L <elastic.port>:<elastic.host>:<elastic.port> sleep <proxy.timeout>
###  ssh -i .ssh/some_id_rsa centos@bastion-host -f -L 9200:remote-es.server.es:9200 sleep 3
# but rust does it for you
proxy.enabled = false
proxy.host = 'proxyhost'
proxy.port = 9201
proxy.protocol = 'http'
proxy.user = 'ec2-user'
proxy.remote_user = 'ec2-remote-user'
proxy.key = 'path to ssh key'
proxy.timeout = 3

Commands Showcase

Learn from examples! Here a few sample commands (and sometimes the output). I hope it is understandable enough.

General

- Get help about the command, options and subcommand

elasticli --help
elasticli <info | index | document> --help

Info

- Get basic info about elasticsearch

elasticli info

{
  "cluster_name": "docker-cluster",
  "cluster_uuid": "b9c-xKsSRs2HQAvZ8wGsIw",
  "name": "aecf011cca00",
  "tagline": "You Know, for Search",
  "version": {
    "build_date": "2023-05-23T17:16:07.179039820Z",
    "build_flavor": "default",
    "build_hash": "c01029875a091076ed42cdb3a41c10b1a9a5a20f",
    "build_snapshot": false,
    "build_type": "docker",
    "lucene_version": "9.6.0",
    "minimum_index_compatibility_version": "7.0.0",
    "minimum_wire_compatibility_version": "7.17.0",
    "number": "8.8.0"
  }
}

Index

- Get index, same operation, multiple ways (it is Clap's magic)

elasticli index -i test1
elasticli index -i=test1
elasticli index --index-name=test1
elasticli index -o read --index-name=test1

- Get indexes ('_all' and '*' are Elasticsearch wildcards)

elasticli index -i _all
elasticli index --index-name='*'

- Create index

elasticli index -o create --index-name='pippo'
elasticli index -o create --index-name='pippo_2' -b '{"settings": { "index": { "number_of_shards": 3, "number_of_replicas": 2 } } }'

{
  "acknowledged": true,
  "index": "pippo",
  "shards_acknowledged": true
}

- Update Index

NOT YET IMPLEMENTED

- Delete Index

elasticli index -o delete -i 'pippo2'
elasticli -c ./samples/default index -o delete --index-name='pippo2'

Document

- Create a document in test1

elasticli document -o create -i test1 -b '{"name":"giufus", "language": "rust"}'

{
  "_id": "5Lms-YgBu6r1vXY7vPX_",
  "_index": "test1",
  "_primary_term": 3,
  "_seq_no": 2,
  "_shards": {
    "failed": 0,
    "successful": 1,
    "total": 2
  },
  "_version": 1,
  "result": "created"
}

- Update an existing document (you need to put your updates in 'doc')

elasticli document -o update -i test1 -b '{"doc": { "language": "zig"} }' --id 5Lms-YgBu6r1vXY7vPX_

{
  "_id": "5Lms-YgBu6r1vXY7vPX_",
  "_index": "test1",
  "_primary_term": 3,
  "_seq_no": 3,
  "_shards": {
    "failed": 0,
    "successful": 1,
    "total": 2
  },
  "_version": 2,
  "result": "updated"
}

- Search all docs in test1 index

elasticli document -o read -i test1

- Search all docs in test1 with an Elasticsearch query

elasticli document -o read -i test1 -b '{ "query": { "term": { "name": "giufus" } }}'

- Delete an existing document

elasticli document -o delete -i test1 --id 5rm3-YgBu6r1vXY7S_Xb

{
  "_id": "5Lms-YgBu6r1vXY7vPX_",
   "_index": "test1",
  "_primary_term": 3,
  "_seq_no": 9,
  "_shards": {
    "failed": 0,
    "successful": 1,
    "total": 2
  },
  "_version": 3,
  "result": "deleted"
}

Defaults

There are some defaults in the code in case you don't specify them:

  • http as protocol
  • 127.0.0.1 as host
  • 9200 as port
  • _doc as type (used in document deletion)
  • read as operation
  • 8.8.0 as es version

To do

  • write better documentation
  • handle elasticsearch versions (e.g. providing multiple implementations of the trait)
  • integration tests (maybe using something like testcontainers)

Run elasticsearch locally with docker and no security

docker run -d --name elasticsearch -p 9200:9200 -p 9300:9300 -e "discovery.type=single-node" -e "xpack.security.enabled=false" elasticsearch:8.8.0

Run elasticsearch locally with docker and basic authentication

docker run -d --name elasticsearch -p 9200:9200 -p 9300:9300 -e "discovery.type=single-node" -e "xpack.security.enabled=true" -e "ELASTIC_PASSWORD=changeme" elasticsearch:8.8.0