rdeer-service 2.2.0

Client-server to handle Reindeer in query mode.

RDEER-SERVICE

rdeer-service is a tool to facilitate the use of Reindeer as a socket mode. It allows multiple Reindeer indexes to be loaded into memory, spread over several servers, enabling queries to be made quickly and easily.

It consists of a server part: rdeer-server, one instance of which will be started per index server, and a client part: rdeer, which allows index manipulation (stopping, starting, etc.) and queries.

It is a companion to Transipedia, a web application for querying Reindeer but it may be useful to use it independently.

Prerequisite

• Reindeer must be installed on indexes servers and the reindeer_socket binary in your $PATH
• You need some Reindeer indexes (Stored on SSD disks to better performances, otherwise build indexes with the Reindeer --mem-query option).

Installation

Recommanded (pip)

python3 -m pip install rdeer-service

Other (git)

git clone https://github.com/Bio2M/rdeer-service.git

How to use?

Start the server

rdeer-server requires Reindeer to be installed on the same physical machine. There can be several instances of rdeer-server running on the same server, each instance listening on a different TCP port (-p option). rdeer-client can query remote servers, so you can have multiple servers hosting rdeer-server/Reindeer and query them all from the same machine.

rdeer-server /path/to/indexes

• rdeer-server listen on port 12800, you can change this with --port option.
• The server will only be able to handle indexes in the specified directory. If your indexes are spread over several directories, you may create symlinks in /path/to/indexes.
• You can add, remove or change the name of the indexes, rdeer-server takes the changes on the fly.
• It is recommended to start rdeer-server as a daemon, using systemd, supervisord or similar.

notifications

rdeer try to restart the indexes cyclically when error stastus

Use the client

The client could requests remote rdeer-server servers. You can enterely manage yours distributed Reindeer indexes with subcommand:

• rdeer list -a to show all indexes with their status
• rdeer start <index-name> to start a index (the index name is the directory hosting the index files)
• rdeer stop <index-name> to stop a index.
• rdeer kill <index-name> If the index crashes during loading, rdeer-server will try to restart. it cyclically. The stop option cannot be used during loading, so the kill option must be used.
• rdeer check <index-name> to verify if index responding.
• rdeer status <index-name> to get the index status (available, loading, running, error).
• rdeer query <index-name> -q <query.fa> to request an index.

show running indexes:

rdeer-client list

Show all indexes handled by rdeer-server

rdeer-client list -a

list all accessible indexes by rdeer-server, with status. Status are :

• available the index is not running
• loading the index is in a transitional mode until the running mode
• running the index is started, and can be resquested.
• error a error occured on the index.

Start an index:

rdeer-client start my-index

Will starts the my-index Reindeer index. When status is running, the index is ready to respond to requests.

Query an index

rdeer-client query my-index -q fasta-query

Requests the specified index, the query file is required and must be in a fasta format.

Options of query subcommand (rdeer-client query --help):

• -q/--query to send a query file at the fasta format (required)
• -f/--format {raw,sum,average,mean,normalize} where
  • raw to get results
  • sum to get sum of kmer counts
  • mean, average to get sum of kmer counts / number of kmers
  • normalize to get normalized counts as billion of kmers
• -s/--server to request rdeer-server on remote host
• -p/--port to request rdeer-server on a specified port (default: 12800)
• -o/--outfile output is stdout by default, you can specified a file