[
https://issues.apache.org/jira/browse/TINKERPOP-2266?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16884754#comment-16884754
]
Divij Vaidya commented on TINKERPOP-2266:
-----------------------------------------
This PR [1] solves this problem by removing the custom keep alive logic and
instead handling the responsibility to Netty [2] . It is not merged in any of
the release branches yet since it carries a lot of other changes with it. Feel
free to take a look at the code and file a PR addressing this specific issue.
Let me know if you need help implementing it.
[1][https://github.com/apache/tinkerpop/pull/1105]
[2][https://github.com/apache/tinkerpop/pull/1105/files#diff-d29dedc9180fcee3c1b87f2f3d24fd4f]
> Keep alive not started at connection creation
> ---------------------------------------------
>
> Key: TINKERPOP-2266
> URL: https://issues.apache.org/jira/browse/TINKERPOP-2266
> Project: TinkerPop
> Issue Type: Bug
> Components: driver
> Affects Versions: 3.3.5
> Reporter: Christian Howe
> Priority: Major
>
> I keep seeing connections in the connection pool being closed in the Gremlin
> Java driver, and it looks like there are no keep alive messages being sent to
> keep the connection open. However, after a write happens to the connection,
> the keep alive seems to start and keep the connection open, based on
> observations from tcpdump. The problem with this is that sometimes when we
> make a query to the client, we get a connection which is closed, and an
> exception is thrown. This results in an increase in customer-impacting
> faults, and retries are likely to pull down another connection which is also
> closed in a pool with a lot of connections. Larger pools are necessary with
> longer running queries to have sufficient concurrency.
> It looks like [when keep alive was
> added|https://github.com/apache/tinkerpop/pull/433], it was written to only
> start the keep alive after there is a write to the connection. In the case
> where a connection is created as part of a connection pool during
> initialization, I can't find where any write would be made to start the keep
> alive. Is there another a mechanism to start this keep alive when a
> connection is created?
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
