[
https://issues.apache.org/jira/browse/ARIA-165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16004303#comment-16004303
]
Ran Ziv commented on ARIA-165:
------------------------------
Several solutions for the node name's suffix have been proposed:
1) Shorter auto-generated IDs - The downside of which is that it's still less
readable, and also that its probabilistic.
2) Using sequential IDs for nodes of a specific node-template (e.g.
{{mynode_1}}, {{mynode_2}}, etc.) - Probably the clearest and most
straightforward way, but it'd require storing additional information on the
node-template's table.
3) Using the node IDs (e.g. {{mynode_20}}, {{mynode_52}}, etc.) - This would
not require any additional information, keep node names unique, and be
relatively straightforward, but might also seem a bit confusing as the IDs
don't really represent the number of nodes a node-template might have. This is
where what Maxim mentioned about SQLAlchemy's events mechanism comes into play
by the way.
4) Finding the maximum suffix of existing nodes and increasing that value by 1
- This is similar to (2), only that it doesn't require storing additional
information - at the price of potentially having nodes with the same name as
nodes which have existed in the past in the same service, which could happen at
the event of a scale down of nodes followed by scaling back up.
As Maxim's mentioned, we've decided to go with (4) at this time. The downside
doesn't seem too terrible since there's always the ability to differentiate
between old and new nodes by their ID, and it allows for readability and
clarity on the likely amount of nodes of a given node-template a service has
had.
> Make node name suffix UUIDs become more readable
> ------------------------------------------------
>
> Key: ARIA-165
> URL: https://issues.apache.org/jira/browse/ARIA-165
> Project: AriaTosca
> Issue Type: Story
> Reporter: Ran Ziv
> Assignee: Maxim Orlov
> Priority: Minor
> Fix For: 0.1.0
>
>
> Node names are currently constructed of the node template name with a UUID
> suffix. While this makes the node name unique across all services, it also
> makes the node names less readable, and since nodes are already unique via
> their ID, it might be better to use a different type of suffix.
> One option would be to generate a shorter, no capital letters string, which
> would be more readable, and (in very high probability) unique in a single
> service.
> Another option would be to go for a running number suffix for each node of a
> given node template, which could be simple, unique and readable.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
