[
https://issues.apache.org/jira/browse/ATLAS-2092?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16150186#comment-16150186
]
Graham Wallis commented on ATLAS-2092:
--------------------------------------
Hi [~ashutoshm],
Thanks for taking the time to read the details. The problem was tricky to find.
I personally haven't seen any commit problems with BerkeleyDB - do you have any
more details about the setup and under what conditions you've seen problems? I
do recall seeing some discussion (maybe in Titan related forums) about
exceptions being swallowed by the graph code and not reported - not sure what
backend was being used in those cases - but just thought I'd mention it here in
case it is affecting what you are seeing.
It sounds like the problem you've had with Berkeley is separate to this
concurrency issue and I'd like to pursue the fix to the concurrency issue first
rather than rebasing just now. I would be interested in setting up HBASE and
Solr though, as it's not a configuration I have used - but not just yet :-)
Your nested transaction patch looks neat. I haven't tried it yet (same reason
as above) but it looks like it would streamline the transactions we perform
against the graph. The other way to approach this might be to not have both the
GraphTransactionInterceptor and the additional transaction wrapper. Could we
remove one of them? My current understanding is that the GTI is the preferred
method, so could the other transaction wrapping be relegated to only being used
in cases that are not annotated as GraphTransaction?
> Failures following concurrent updates
> -------------------------------------
>
> Key: ATLAS-2092
> URL: https://issues.apache.org/jira/browse/ATLAS-2092
> Project: Atlas
> Issue Type: Bug
> Components: atlas-core
> Reporter: Graham Wallis
> Attachments: import-nested-txn.patch, Investigations and findings
> relating to concurrent updates in Atlas.pdf
>
>
> There is a race condition that causes duplication of schema vertices as a
> result of concurrent graph updates. This in turn leads to failure of queries
> that specify a type such as an edge label used in an attribute that
> references another entity. This problem is known to affect Atlas entity refs
> – which create graph edges that use edge label schema vertices. It is likely
> that it also affects other types in Atlas.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
