Skip to content

Latest commit

 

History

History

sentinel

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
redis_configs
redis_configs
 
 
README.md
README.md
 
 
installation.sh
installation.sh
 
 
model_run.py
model_run.py
 
 
model_set.py
model_set.py
 
 
run_server.sh
run_server.sh
 
 
setup.py
setup.py
 
 

README.md

Setting up Sentinel

The whole configuration process is scripted in this folder for setting up three node Redis server and three node sentinel instance. We are setting up one sentinel and one Redis-server in one node and hence need three nodes.

  • SSH to the server and run installation.sh: This will set up the Redis-server, RedisAI, and prerequisites required. This step installs RedisAI for CPU, if you need to try out GPU based sentinel setup, you might have to make a slight tweak in the script.
  • Execute setup.py as python setup.py --master "<master IP>" --slaves "<comma separated slave IPs>" --passphrase <ssh key passphrase> --user <username> (Make sure you have fabric installed). This copies the configuration (redis.conf and sentinel.conf) to remote machines. The script we made assumes home folder in all the machines are the same (username is the same for all machines)
  • Run run_server.sh in each machine to bring up the Redis master and slave instances with RedisAI loaded.
  • For testing: run model_set.py on the master which sets the imagenet - pytorch model on the master instance. Then go ahead and bring down the master instance. This makes the sentinels elect the new master. Now run model_run.py (old master is still down) to get the model result.