directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ole Ersoy (JIRA)" <j...@apache.org>
Subject [jira] Commented: (DIRSERVER-950) Server Left in Unstable State
Date Sat, 02 Jun 2007 21:02:15 GMT

    [ https://issues.apache.org/jira/browse/DIRSERVER-950?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12500968
] 

Ole Ersoy commented on DIRSERVER-950:
-------------------------------------

Just wanted to make sure I clearly explain what the challenge/concern is.  My main concern
is that the server is left in a state that cannot be recovered from.  This test shows that
this is a fact:
https://issues.apache.org/jira/browse/DIRSERVER-950#action_12500469

My attempt to isolate why this was happening lead me to these observations.  If an attempt
to delete an entry that other entries depend on, the server tries to delete the entry (Only
partly succeeds), instead of throwing an exception.  This in turn to leads to the irrecoverable
state.  This is my hypothesis about the cause, and the reason that I think it is true is because
I can delete the same two ObjectClasses in the proper order, and the server is fine.  In other
words if I delete 1 then 2, the server is fine, but if I delete 2 then 1, I cannot clean these
entries anymore (They are partially deleted), and their parent context cannot be removed.
 I have to reinstall the server if I want these to be cleaned.

> Server Left in Unstable State
> -----------------------------
>
>                 Key: DIRSERVER-950
>                 URL: https://issues.apache.org/jira/browse/DIRSERVER-950
>             Project: Directory ApacheDS
>          Issue Type: Bug
>    Affects Versions: 1.5.0
>         Environment: FC6
>            Reporter: Ole Ersoy
>            Priority: Minor
>
> I can only give observational data
> for this, since it appears to be a one time thing.
> It's a little like the sequence of what happened depends
> on threads, and thread A completes first 9999/10000 times,
> but in this case thread B completed first...
> I ran the test:
> EObjectClassCreator
> It creates an ObjectClass and the 
> AttributeTypes needed by the ObjectClass.
> I've ran it several times successfully already,
> but on the last run it left the server in 
> a state that does not let me delete
> an entry shown in LS.
> I have a single entry (That should have been cleaned by the test)
> ou=objectClasses, cn=ecore, ou=schema
> This has no children.
> If I refresh in LS it is still there.
> I check to make sure the apacheds daemon is still running.
> If I try to delete I get this:
> Error while deleting entry
> [LDAP: error code 80 - failed on search operation: OID for name 'metatopsdo' was not
found within the OID registry]
>   [LDAP: error code 80 - failed on search operation: OID for name 'metatopsdo' was not
found within the OID registry]
> metatopsdo is an objectClass that was created in order to create 
> the ObjectClass entry that the test creates.
> Once the test has run the ObjectClass entry
> created is deleted,
> then the metatopsdo ObjectClass.
> This is strange because the entry
> ou=objectClasses, cn=ecore, ou=schema
> is empty, at least as shown is LS.
> So the question is under what circumstances does
> is LS not able to delete an empty entry like this.
> I'm now going to disconnect and reconnect
> with LS to see if it has an effect.
> OK - I did that and
> when drill down past the entry
> ou=objectClasses, cn=ecore, ou=schema
> I get this:
> Error while reading entry
> [LDAP: error code 80 - failed on search operation: OID for name 'metatopsdo' was not
found within the OID registry]
>   [LDAP: error code 80 - failed on search operation: OID for name 'metatopsdo' was not
found within the OID registry]
> So the ObjectClass entry the test created
> must be there, but LS can't see it because
> the metatopsdo ObjectClass is deleted,
> at least it seems.
> But ApacheDS should disallow deletion
> of metatopsdo when it is being used by another 
> ObjectClass entry right?
> I think the only thing I can do now is to
> reinstall the server...since the entry
> is "Stuck"...
> But first let me try restarting it 
> just in case...
> yum still stuck.
> Just in case the teardown for this test looks like this:
>         EObjectClassDestroyer.
>         destroy( 
>             ecoreAttributeTypesContext, 
>             ecoreObjectClassesContext, 
>             eClass, 
>             oidPrefix );
>         MetaTopSDOObjectClassDestroyer.
>         destroy(
>             ecoreAttributeTypesContext, 
>             ecoreObjectClassesContext, 
>             oidPrefix );        
>         
>         EcoreTypeSystemDestroyer.
>         destroy(
>             ecoreSyntaxesContext, 
>             oidPrefix);
>         
>         super.tearDown();
> So it's really strange that it got stuck like this...

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message