[
https://issues.apache.org/jira/browse/IGNITE-9171?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16728745#comment-16728745
]
Taras Ledkov commented on IGNITE-9171:
--------------------------------------
[~vozerov],
1. Connection manager changes was corresponded to previous lazy impl: with lazy
threads. Your suggestion make sense.
But current connection manager has *potential connection leak*:
detached connection is removed from the map {{threadConns : Thread->
H2ConnectionWrapper}} and there is no way to close all connection on the node
stop.
2. Rolls back changes at the connection manager highlight the problem with
detached connection on reducer. Looks like it was cause of the error. I'll
check it on TC,
> Use lazy mode with results pre-fetch
> ------------------------------------
>
> Key: IGNITE-9171
> URL: https://issues.apache.org/jira/browse/IGNITE-9171
> Project: Ignite
> Issue Type: Task
> Components: sql
> Affects Versions: 2.6
> Reporter: Taras Ledkov
> Assignee: Taras Ledkov
> Priority: Blocker
> Labels: sql-stability
> Fix For: 2.8
>
>
> Current implementation of the {{lazy}} mode always starts separate thread for
> {{MapQueryLazyWorker}}. It causes excessive overhead for requests that
> produces small results set.
> We have to begin execute query at the {{QUERY_POOL}} thread pool and fetch
> first page of the results. In case results set is bigger than one page
> {{MapQueryLazyWorker}} is started and link with {{MapNodeResults}} to handle
> next pages lazy.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
