topic Re: Solr index failing (corrupt data?) in Alfresco Archive

Solr index failing (corrupt data?)

kriton — Tue, 26 Jun 2012 12:04:49 GMT

Solr Indexing on my live environment has stopped working for a couple of weeks now. Any content I have I added since the 8th of June is not tracked by Solr anymore and I cannot figure out why.This is an exception that repeats itself in the log every few minutes:]] DB row: NodeEntity[ ID=30163,

Re: Solr index failing (corrupt data?)

andy — Tue, 26 Jun 2012 16:45:49 GMT

Hi

SOLR can skip this - the normal tracking is unforgiving

See the FIX action on the wiki page - it will inedx up to date(ish) and skip any problem transactions.
You can retry by re-runnning FIX.

Is the cache issue resolved by restarting alfresco?
Are you running in a cluster? Have you got the cache configured correctly?
It may be worth trying this first.

Andy

Re: Solr index failing (corrupt data?)

kriton — Thu, 28 Jun 2012 12:45:16 GMT

Hi Andy, thanks for the reply.

The FIX action I have tried and does absolutely nothing. I have also tried the PURGE actions and they dont seem to work either….

I am not running in a cluster and restarttng alfresco doesnt change anything.

It seems that solr tracker has fallen into some sort of a loop. It tries to read a node and retrieve information about it. It fails to do that, then after a few seconds it retries for the same node, etc.

I have identified a call made from SOLR to Alfresco to retrieve metadata information. I simulated this call for all nodes in the database and I identified that 4 nodes fail and return 500 status code. The first of them is the one that causes SOLR index to loop.

The webscript that is called from within SOLR is /api/solr/metadata

I am trying to identify what these nodes may have that is faulty and fix it. Any ideas ?

thanks again

Re: Solr index failing (corrupt data?)

andy — Thu, 28 Jun 2012 14:06:29 GMT

Hi

Your error suggests that a deleted node is ending up in the Node cache and then it is not live.
Have you upgraded? What version of Alfresco did you start with?

There is a job to remove deleted nodes.

Set
index.tracking.minRecordPurgeAgeDays=0
to remove all deleted nodes

I have not checked how often this is scheduled to run.

(The default is to keep deleted nodes for 30 days to support index tracking)

Andy

Re: Solr index failing (corrupt data?)

kriton — Thu, 28 Jun 2012 15:01:50 GMT

I started with 4.0.b and a month ago upgraded to 4.0.d

3 of those nodes were thumbnails that were refering to node 30163 as their parent. I marked them as deleted in the database and problem gone.

My questions:
- Any idea how this could have happaned in the first place ?

- The job that you told me about, is this supposed to do the same thing that I manally did in the database ? And how can I actually set it to run on demand ?

Thanks again

Re: Solr index failing (corrupt data?)

andy — Fri, 29 Jun 2012 18:44:23 GMT

Hi

The job removes deleted nodes.

The issue you had was live thumbnails pointing to a deleted doc and the tracker failing to get its path as a result.

at org.alfresco.repo.domain.node.AbstractNodeDAOImpl.getParentAssocsCached(AbstractNodeDAOImpl.java:3586)
This is a known (and fixed) issue.

The casue is fixed.
SOLR will also skip any issue like this and log an exception in the index.

You will need an Enterprise build for the latest fixes, pick it up when they are merged into the nightly build - along with lots of other stuff,
or wait for the next official community release.

Andy

Re: Solr index failing (corrupt data?)

kriton — Tue, 03 Jul 2012 09:24:58 GMT

thanks a lot for your replies Andy

Re: Solr index failing (corrupt data?)

bhramari — Wed, 17 Dec 2014 08:37:11 GMT

Hi,

I was trying to re-index my alfresco solr, but the index doesn't build up and everytime it shows only one file inside the index folder "lucene-e79c4944314431e4e034db76c030ec86-lucene-2vfwnv-test.lock"

Please help!
its very urgent!