[
https://issues.apache.org/jira/browse/IGNITE-11258?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16784215#comment-16784215
]
Vladimir Ozerov commented on IGNITE-11258:
------------------------------------------
[~alapin], my comments:
# {{JdbcPinnedResult}} - and again, this is not {{pinned}} result, so class
name is misleading. This is just a result with IO it was sent through. Let's
just name it {{JdbcResultWithIo}} then?
# Let's move {{HandshakeResult}} to a top-level. Inner classes are typically
lead to poor encapsulation and hidden bugs as a result.
# {{JdbcThinStatement}} - please rename {{stickyIo}} local variables and
parameters when they are not really {{sticky}}, but instead just denotes the IO
some request was sent through. We need to have clear naming to simplify future
development which will be performed by possibly other contributors.
# {{JdbcThinConnection}} - let's review naming as well. We have a lot "ios"
with no systematic approach to their naming:
## {{cliIo}} -> {{singleIo}]
## {{nodeToConnMap}} -> {{ios}}
## {{connections}} -> {{iosArr}}, and why is it {{Object[]}} instead of
{{JdbcThinTcpIo[]}}?
## {{txStickyIo}} -> {{txIo}}
# {{JdbcThinConnection.sendRequest}} - {{ownThread = null}} should be done
inside separate {{finally}} block to be on the safe side.
# {{connectInCommonMode}} and {{connectInBestEffortAffinityMode}} - there is
code duplication in exception handling. This should be factored out.
> JDBC Thin: update connection setup logic.
> -----------------------------------------
>
> Key: IGNITE-11258
> URL: https://issues.apache.org/jira/browse/IGNITE-11258
> Project: Ignite
> Issue Type: Task
> Components: jdbc
> Reporter: Alexander Lapin
> Assignee: Alexander Lapin
> Priority: Major
> Labels: iep-23
> Fix For: 2.8
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> # On thin client startup it connects to *all* *nodes* provided by user by
> client configuration.
> # Upon handshake server returns its UUID to client.
> # By the end of the startup procedure, client have open connections to all
> available server nodes and the following mapping (*nodeMap*): [UUID =>
> Connection].
> Connection to all nodes helps to identify available nodes, but can lead to
> significant delay, when thin client is used on a large cluster with a long IP
> list provided by user. To lower this delay, asynchronous establishment of
> connections can be used.
> For more information see [IEP-23: Best Effort
> Affinity|https://cwiki.apache.org/confluence/display/IGNITE/IEP-23%3A+Best+Effort+Affinity+for+thin+clients]
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
