ashrithr / ankus Goto Github PK

Currently mapreduce deployment cannot be turned off, but this should be possible with hbase deployments which performs better with out mapreduce on the cluster.

AnkusCLI doesn't work on ruby 1.8.7

clear the .data/* files for cloud deployments if the cluster is destroyed

Update puppet_installer to support puppet 4.0.

If Deployment mode is cloud and cloud_platform is rackspace, ankuscli destroy subcommand will delete previously created servers by ankuscli because of the way ankuscli is storing the instances in cloud_instances.yaml by their hostnames instead try to use id of the instances as well and destroy based on that.

Racksapce api tends to take more time creating the instances, so if for some reason if rackspace cannot create instances with in specified time, safely delete create instances and exit out

--debug mode adds extra tabs and new lines for some reason!

Right now, building a VM expanding the root volume and mounting & formatting volumes if any takes forever to complete, built pre-built amazon ami's with already formatted and mounted volumes(ebs)

Also, puppet master installation (with passenger, puppetdb) takes sometime which can be cut-down with a pre-built ami as well.

Deleting instances with --delete-volumes for aws deployments is blocking make it multithreaded so that would consume far less time to destroy

This is a test

Problem doing ssh into instances created using rackspace cloud_platform, because rackspace can only take public_key to create instances.

if user passes --delete-volumes to destroy subcommand detach and delete attached block store volumes

If a user changes the configuration files, should reload the re-run puppet on all nodes reflecting the changes back on the cluster.

For some reason, puppetdb cannot be initialized in rackspace based deployments

• info subcommand should support cluster wise informations with roles and modular cluster info (ankus cluster info should be optional).
• support to change the ankus assigned roles for nodes deployed in cloud
• add --reload for deploy subcommand to deprecate --add-nodes, --hosts, --count, --roles options, which should be able to update the deployment based on the config changes.

Write a MCollective plugin to perform the orchestration.

Ankus should natively support container based deployment's on top of docker like frameworks.

ConfigParser class should parse the configuration file and show errors and warnings for all the parameters and then exit out raising Ankuscli::Errors::ParseError

Test hadoop & hbase deployments with ubuntu

currently local deployments mode dont support ssh'ing using ssh subcommand.

Do somkind of auto-discovery and identify volumes and their mount points and point hdfs.data.dir to them.

right now ssh_user is prefixed as root, user should be able to pass any user who has passwordless sudo privilges to all servers.

add-node should add new node to the existing cluster, this sub-command should take in option --type which specifies what kind of node is being added to the cluster.

For ex: --type hadoop should provision new node with roles datanode and tasktracker
--type hbase should provision new node with role region server
if option is not specified then provision node based on the configuration file

Rack space api tends to fail while creating servers if we caught an error during build phase. Fail safely by deleting created servers and raised timeout.

Integrate Google cloud engine for cloud deployments

ankus should support cassandra deployments

There are no status updates even if '--debug' is passed, add some kind of progress indicator when installing or running puppet.

ganglia works with local and rackspace deployments but cannot get it to work with aws deployments, it has to do something with aws security groups