Mailing-List: contact issues-help@geode.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@geode.incubator.apache.org
Date: Fri, 8 Apr 2016 16:27:25 +0000 (UTC)
From: "ASF subversion and git services (JIRA)" <jira@apache.org>
To: issues@geode.incubator.apache.org
Message-ID: <JIRA.12956276.1459887981000.176357.1460132845586@Atlassian.JIRA>
In-Reply-To: <JIRA.12956276.1459887981000@Atlassian.JIRA>
References: <JIRA.12956276.1459887981000@Atlassian.JIRA>
 <JIRA.12956276.1459887981305@arcas>
Subject: [jira] [Commented] (GEODE-1177) Session replication local cache is
 reporting bad object sizes
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/GEODE-1177?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15232434#comment-15232434 ] 

ASF subversion and git services commented on GEODE-1177:
--------------------------------------------------------

Commit e0c20de1645d65bd67eb61350c2ad474b341998c in incubator-geode's branch refs/heads/feature/GEODE-17-2 from [~upthewaterspout]
[ https://git-wip-us.apache.org/repos/asf?p=incubator-geode.git;h=e0c20de ]

GEODE-1177: Adding a few more classes to exclude from object sizing

This is not a complete fix for the http session management problem, but
this is a safety catch we should probably have in the product to avoid
following references into InternalDistributedSystem, Classloaders, or
Loggers when computing the size of a user's object.


> Session replication local cache is reporting bad object sizes
> -------------------------------------------------------------
>
>                 Key: GEODE-1177
>                 URL: https://issues.apache.org/jira/browse/GEODE-1177
>             Project: Geode
>          Issue Type: Bug
>          Components: http session
>            Reporter: Dan Smith
>
> The session replication module has a local cache that uses heap LRU - See PeerToPeerSessionCache.createOrRetrieveLocalRegion.
> When using LRU, the region uses an object sizer to determine the size of the objects. The default object sizer finds all reachable objects using reflection for the first value in the region, and then reuses that size for all subsequent values of the same class. See EvictionAction.createLRUHeapAttributes().
> In the case of session replication, it looks like we put a GemfireHttpSession in the region. Unfortunately, that has references back to the session manager, which has references to an InternalDistributedSystem, which in turn pulls in a whole lot of junk.
> So the size of the objects reported by the sizer for this local cache are probably completely wrong, which might affect the behavior of the heap LRU.


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)