Skip to content

A multi-dimensional monitoring frame-work for geo-distributed applications using heterogeneous hardware

Notifications You must be signed in to change notification settings

NolanRudolph/MicroMon

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

85 Commits
 
 
 
 
 
 
 
 

Repository files navigation

SETUP

This file explains how to setup my project. Please follow the CLIENT or SERVER & REPLICAS section respective of the node you're setting up. Both types of nodes should begin by following the BOTH section.

BOTH

  1. cd ~
  2. git checkout https://github.com/SudarsunKannan/butterflyeffect.git
    a. Provide proper credentials
  3. cd butterflyeffect
  4. git checkout FullDynamic
  5. cd code

CLIENT

  1. Run "bash scripts/client_setup.sh"
  2. Change the HOST variable in ~/butterflyeffect/code/scripts/setvars.sh to the IP of the host node in the Cassandra cluster. Then run "source scripts/setvars.sh" from ~/butterflyeffect/code.

SERVER & REPLICAS

  1. Make sure you follow REQS.txt found in the root directory of this repository
  2. From ~/butterflyeffect/code, run "bash scripts/replica_setup.sh"
  3. On each replica node and the server, you must change some configuration files found in ~/butterflyeffect/code/cassandra/conf.
    a. Change cassandra.yaml so that
          1. seeds on line 428 includes the IP addresses of all nodes in the cluster with a comma delimiter
          2. listen_address on line 615 is the IP address of this node
          3. rpc_address on line 692 is the IP address of this node
    b. Only on server, change cassandra-topology.properties so that they replicate a format identical to lines 18-20; That is,     IP=DC#:RAC#. An entry like this should be created for each node found within "seeds" from line 428 of cassandra.yaml.     I.e., create an entry for each node in the cluster, including itself.
    c. Change cassandra-rackdc.properties so that
          1. dc=dc# on line 19, where # reflects the same datacenter as its configuration in the server's
              cassandra-topology.properties
          2. rack=rack# on line 20, where # reflects the same rack as its configuration in the server's
              cassandra-topology.properties

RUNNING TESTS

  1. Make sure that setvars.sh has been sourced from ~/butterflyeffect/code. You can check this by calling "echo $CODE" and making sure it equates to ~/butterflyeffect/code.
  2. Make sure that the cassandra.yaml, casssandra-rackdc.properties, and cassandra-topology.properties (on the server) have all been updated to hold the correct content (see SERVERS & REPLICAS step 3)
  3. On the server node and all replica nodes, from ~/butterflyeffect/code, as simultaneously as possible call "bash scripts/restart.sh". This kills any Cassandra processes, deletes the Cassandra data folders, and restarts Cassandra. Therefore, this works as the name suggests: A restart from a previous test, but it also works perfectly fine for a lone test as well.
  4. Wait until you see "No gossip backlog; proceeding" from the Cassandra logger on all nodes.
  5. Now, on the client node, make sure "echo $HOST" returns the IP of the server. If it does not, edit the HOST variable in ~/butterflyeffect/code/scripts/setvars.sh, and then run "source scripts/setvars.sh" from ~/butterflyeffect/code (see CLIENT step 2)
  6. On all nodes, run "bash ~/butterflyeffect/code/scripts/flush.sh"
  7. On the client node, make sure line 4, 23, and 32 of ~/butterflyeffect/codescripts/customTest.sh are updated to your liking. Line 4 holds the variable "OPSCNT", which is the number of entries that will be loaded during the load phase of Cassandra. The load phase of cassandra is executed on line 23 - To change workloads, change "workloada" to "workloadX", replacing X with a letter from [a, b, c, d, e]. The same explanation applies to line 32. You can inspect and manipulate what these different workloads entail by viewing ~/butterflyeffect/code/mapkeeper/ycsb/YCSB/workloads/workloadX, again with X being replaced with a letter from [a, b, c, d, e].
  8. From ~/butterflyeffect/code, run "bash scripts/customTest.sh".
  9. IMPORTANT: Between the load phase and run phase, make sure to run "bash ~/butterflyeffect/code/scripts/flush.sh" on all nodes but the script node.

SWITCHING TO VANILLA

Note: This will be followed on all nodes except Client

  1. Run "cd ~/vanilla/code"
  2. Run "source scripts/setvars.sh"
  3. If you have followed step 3 from SERVER & REPLICA, run the following commands:
    a. "cp ~/butterflyeffect/code/cassandra/conf/cassandra.yaml ~/vanilla/code/cassandra/conf/cassandra.yaml"
    b. "cp ~/butterflyeffect/code/cassandra/conf/cassandra-rackdc.yaml ~/vanilla/code/cassandra/conf/cassandra-rackdc.properties"
    c. On server, "cp ~/butterflyeffect/code/cassandra/conf/cassandra-topology.properties ~/vanilla/code/cassandra/conf/cassandra-topology.properties"
    d. "vim ~/vanilla/code/cassandra/conf/cassandra.yaml", and in this file, type ":%s/butterflyeffect/vanilla/g", then ":wq"
  4. If you have not followed step 3 from SERVER & REPLICA, I strongly encourage you to do this first. If not, follow step 3 from SERVER & REPLICA, except replace all instances of "butterflyeffect" with "vanilla"

About

A multi-dimensional monitoring frame-work for geo-distributed applications using heterogeneous hardware

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published