[#125] - Enhancing clustering functionality #238
Conversation
Transfering rabbitmq-cluster cookbook into rabbitmq cookbook
* Changing attribute prefix : [:rabbitmq-cluster] to [:rabbitmq][:clustering]
* Enhancement on regex : user ruby matcher
* Removing unnecessary resource : node_type on change_cluster_node_type action
* Some other enhancement on provider
Enhancing from rabbitmq-cluster as the manual clustering
* Removing master/slave concept
* Removing cluster node type
* Introducing new attribute that defines all the cluster nodes and its cluster node type : `[:rabbitmq][:clustering][:cluster_nodes]`
* Adding set_cluster_name action
Applying new features and attributes into auto clustring
* Using `default[:rabbitmq][:clustering][:cluster_nodes]` attribute instead of `['rabbitmq']['cluster_disk_nodes']`
* Applying set_cluster_name and change_cluster_node_type for auto clustering
* Long story short, All the nodes will be set as `disc` node in the rabbitmq.config even though you have some ram nodes. however, change_cluster_node_type action will fix the node type as you expected.
Additional changes on the others
* Fixing typo in default receipe for stopping service to set erlang cookie
Attributes example
```
default[:rabbitmq][:clustering][:use_auto_clustering] = false
default[:rabbitmq][:clustering][:cluster_name] = 'seoul_tokyo_newyork'
default[:rabbitmq][:clustering][:cluster_nodes] = [
{
:name => 'rabbit@rabbit1',
:type => 'disc'
},
{
:name => 'rabbit@rabbit2',
:type => 'ram'
},
{
:name => 'rabbit@rabbit3',
:type => 'disc'
}
]
```
Redesign of manual clustering
- remove master/slave
- remove cluster node type for each node
- introducing new attribute that defines all the cluster nodes and its cluster node type
- adding set_cluster_name feature
- preparing of auto clustering in cluster recipe
Adding set cluster name action for manual clustering.
Apply [:rabbitmq][:clustering][:cluster_nodes] attribute on rabbitmq.config
fix typo
Apply set_cluster_name and change_cluster_node_type for auto clustering
|
Thank you for taking the time to work on this. I am taking a look through it now. Can you fix the style issues that travis is complaining about? Also I think we might want to a section to the README describing how to use these changes. |
|
@cmluciano |
- Fix some code to handle style issue.
|
@cmluciano I fixed style issue and rubocop issue. and Travis has passed. please review it. |
| # execute > rabbitmqctl cluster_status | sed "1d" | tr "\n" " " | tr -d " " | ||
| # rabbitmqctl cluster_status returns "Cluster status of node rabbit@rabbit1 ..." at the first line. | ||
| # To parse the result string, it is removed by sed "1d" | ||
| cmd = 'rabbitmqctl cluster_status | sed "1d" | tr "\n" " " | tr -d " "' |
There was a problem hiding this comment.
I'm a little confused by the need to pipe to sed, tr. etc. Is this to try to normalize the output?
There was a problem hiding this comment.
Yeah i have to parrot @cmluciano here, this cmd scares me. Is there a way we could leverage some ruby or something other then sed tr tr?
There was a problem hiding this comment.
Yeah, that is for normalize the output. because, rabbitmqctl output, mostly indent and line feed, was different from servers and versions.
There was a problem hiding this comment.
I'l try with ruby.
There was a problem hiding this comment.
It would be changed like this.
def cluster_status
# execute > rabbitmqctl cluster_status"
# To parse the result string, this function normalize the output string
# - Removing first line
# - Removing "... Done" : old version returns this
cmd = 'rabbitmqctl cluster_status'
Chef::Log.debug("[rabbitmq_cluster] Executing #{cmd}")
cmd = get_shellout(cmd)
cmd.run_command
cmd.error!
result = cmd.stdout.split(/\n/, 2).last.squeeze(' ').gsub(/\n/, '').gsub('...done.', '')
Chef::Log.debug("[rabbitmq_cluster] rabbitmqctl cluster_status : #{result}")
result
end
- Using ruby code instead of linux command : `sed`, `tr` has removed. - Fixing english grammar - Removing `json` tag from README : its not from the code review result
|
@jjasghar @cmluciano |
|
I restarted it, lets see if it passes. |
|
@jjasghar Looks good! |
|
👍 |
| # Join in cluster | ||
| rabbitmq_cluster cluster_nodes do | ||
| action :join | ||
| end |
There was a problem hiding this comment.
Should this unless block be a not_if {} in the resource call instead?
There was a problem hiding this comment.
@jesseadams
node['rabbitmq']['clustering']['use_auto_clustering'] is the flag for using auto clustering or manual clustering. should I use not_if? can you please explain to me how not_if different from unless at this code?
There was a problem hiding this comment.
It is fundamentally the same. I thought that not using not_if was a FoodCritic violation in this case. Maybe not?
|
Can we please get this reviewed once more and merged? Thanks! |
Transfering rabbitmq-cluster cookbook into rabbitmq cookbook
Enhancing from rabbitmq-cluster as the manual clustering
[:rabbitmq][:clustering][:cluster_nodes]Applying new features and attributes into auto clustring
default[:rabbitmq][:clustering][:cluster_nodes]attribute instead of['rabbitmq']['cluster_disk_nodes']discnode in the rabbitmq.config even though you have some ram nodes. however, change_cluster_node_type action will fix the node type as you expected.Additional changes on the others
Attributes example