[
https://issues.apache.org/jira/browse/DERBY-6985?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16382092#comment-16382092
]
Satyabrata Mohanty commented on DERBY-6985:
-------------------------------------------
Hi Rick,
Thanks for the information. Its now clear that this is configured in OS level.
Now one important point for us is, if any scenario happens in production as
mentioned in issue description section, where the dead connection keep
increasing due to network reset, then there might be large number of
connections (as dead which will be cleared after default limit ) and also there
will be new connections in derby server when network is up.
So what is the threshold limit of number of opened connection in derby, is
there any property parameter where we can set the maximum limit of connection
that derby server will handle or it depends upon heap memory size of the system
where derby server is running.
For analyzing threshold, we made an java application which is continuously
creating connection without closing, below are some additional details,
Derby Server System RAM - 32 GB
Derby Allocated Heap memory - 12 GB
When open connection count on derby server is *3966*, we got below exception,
Caused by: java.io.FileNotFoundException:
/usr/java/jdk1.8.0_65/jre/lib/currency.data (Too many open files)
at java.io.FileInputStream.open0(Native Method)
at java.io.FileInputStream.open(FileInputStream.java:195)
at java.io.FileInputStream.<init>(FileInputStream.java:138)
at java.io.FileInputStream.<init>(FileInputStream.java:93)
at java.util.Currency$1.run(Currency.java:221)
... 39 mores
We need this information regarding the threshold limit of derby server as this
scenario may arise in production where derby server might crash due to high
number of opened connection generated due to network reset like scenario.
Please provide us the information on this and also let us know for any
information if needed.
Thanks,
Satyabrata
> Document how to configure the keepAlive timeout on network sockets
> ------------------------------------------------------------------
>
> Key: DERBY-6985
> URL: https://issues.apache.org/jira/browse/DERBY-6985
> Project: Derby
> Issue Type: Bug
> Components: Documentation
> Affects Versions: 10.14.1.0
> Reporter: Satyabrata Mohanty
> Priority: Major
> Attachments: Derby_Open_Connection.png
>
>
> Hi,
> Below is the complete description of the issue,
> * Running a java application which creates a good number of connections with
> derby on the machine X.
> * We are running the derby database server on the machine lets say Y.
> * We explicitly block the listening of 1527 port of system X traffics on
> system Y in the network until it thrown connection exception and then again
> established the network between the two systems.
> * After the java application has finished execution we found there are still
> some connections remained open in derby.
> * We can see open connections in derby although we have terminated the java
> application.
> * We have restarted derby to free up these connection.
> * Concern here is, how these connection are generated, is these connections
> are broken /stale connection made during the network reset, if YES then, how
> to these connection will be auto deleted or handled in derby ?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
