deployment.md

Deployment

The current deployment uses the existing VM configuration and basically is
(a) downloading the current version from GitHub,
(b) creating a link to /opt/config.py and
(c) linking from /opt/embeddings.

• Login
  • kinit wilke
    (or another upb user)
  • ssh wilke@embeddings.cs.upb.de
• Source code to /opt
  • emb=/opt/embeddings.cc-2022-06-30
    (or another directory)
  • cd /tmp
  • wget https://github.com/dice-group/embeddings.cc/archive/refs/heads/master.zip
  • unzip master.zip ; rm master.zip
  • mv embeddings.cc-master $emb
• Link config
  • cd $emb
  • mkdir instance
  • ln -s /opt/config.py instance/
• Link new instance
  • sudo systemctl stop embeddings.service
  • unlink /opt/embeddings
  • ln -s $emb /opt/embeddings
  • sudo systemctl start embeddings.service
• Eventually delete old deloyment directories

Starting the index webservice

• screen -S webservice-index
• . /opt/bashrc.sh
• cd /opt/embeddings
• ./scripts/run-webservice-index-vm.sh
• Test: http://embeddings.cs.uni-paderborn.de:8008/ping

VM config overview

• Elasticsearch API
  • ES :9200 version
  • ES :9200 indexes
  • ES :9200 mappings example
• Elasticsearch 8.3.1
  • Manual start command: see Guide for starting, stopping)
  • Check ES logs command: tail /data/es8-logs/embcc.log
• Debug webservice-public
  • Use run-webservice-public-local.sh to start another instance
  • Access at http://embeddings.cs.upb.de:1337

Archive

Use deployment commands in screen session. Before, activate environment:

. /opt/anaconda3/etc/profile.d/conda.sh
conda activate embeddings