[
https://issues.apache.org/jira/browse/IGNITE-790?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14511349#comment-14511349
]
Denis Magda commented on IGNITE-790:
------------------------------------
Implemented using jcloud.
Everything works but performance is unacceptable. Depending on instances count
it can take from several seconds to several minutes to get a list of registered
nodes using jcloud API. Seems that jcloud SSHes to every instance while
gathering node's info. No way to make a butch operation in jcloud. Will
investigate jcloud sources and if there is no way to avoid SSH then we can't
use jcloud for this task.
> Cross-cloud IpFinder for nodes auto discovering
> -----------------------------------------------
>
> Key: IGNITE-790
> URL: https://issues.apache.org/jira/browse/IGNITE-790
> Project: Ignite
> Issue Type: Improvement
> Components: general
> Reporter: Denis Magda
> Assignee: Denis Magda
> Fix For: sprint-5
>
>
> The goal is to create an IpFinder that can be used with many cloud providers
> without a necessity to create a vendor specific implementation.
> The IpFinder must discover Ignite nodes automatically.
> At least at the first iteration the workflow is the following:
> - the finder gets private and public (?) IP addresses (without port numbers)
> of all cloud machines using corresponding jcloud API;
> - optionally the addresses may be filtered out by some predicate (elaborate
> lately);
> - the addresses returned in getRegisteredAddresses() as is, the underlying
> code will add a valid port number;
> User is responsible for:
> - setting all required cloud specific parameters (credentials, secret key);
> - setting a port number that will be used by every Ignite node on every
> machine.
>
> Limitations:
> - only one Ignite node will run per a cloud virtual machine.
> When done, existed S3 and Google Cloud Storage implementations should be
> deprecated.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
