Indexing fails with CCK nodereference fields

Garrett, can you upgrade to DRUPAL-5 and try that (the devel release)? Please confirm the bug with that code.

Searching for *:* will give you all results in the index.

Solr is very flexible in what you can send its way. It only has one required field, and that is nid. Your documents can have any other fields they want, and you can omit any of the other fields. Somewhere your site is trying to index something that doesn't have an nid, so debugging this will require a bit of sleuth work on your part.

At line 131 in apachesolr.module you'll see this:
$node = node_load($row->nid);
Right after that, add this:

if (empty($node->nid)) {
  watchdog('apachesolr', "Looked for " . $row->nid . " found this:" . print_r($node, true), WATCHDOG_ERROR);
}

Then reproduce the error and inspect your watchdog messages.

Let me know what you find.

In the meantime I added a check to the latest DRUPAL-5 version that should just avoid the problem by skipping nodes that don't load.

do you have some optional nodereference cck fields somewhere?

i enabled some nodereference cck fields inside a field group and i think this started to give me exactly the same errors as you are receiving

I have a similar problem. Here are the key bits of information I can provide:

• I see similar taxonomy errors. However, these errors occur whether or not apachesolr is enabled, making me doubt apachesolr is the root cause.
• apachesolr throws an exception every time it tries to index a document. The message provided is "0" Status: Communication Error. This seems to be the default Solr error.
• apachesolr pings the Solr instance without difficulty, so the basic configuration seems right.
• my tomcat logs show the ping requests from Drupal, but no other request is ever logged. I have increased tomcat's logging level and still get nothing. I suspect the requests die in the SolrPhpClient and are not sent to Solr.

I am versed in Solr but less so with Drupal. I have used serialize() and print_r to examine the documents apachesolr tries to index, but don't know what the results should look like.

I would really like to try this module and would be happy to debug further, I'm just not sure where to start. I'm open to suggestions...

thanks!
-Graham

We obviously have to fix this before a 1.0 release.

I'm taking a look at tackling this issue but have not been able re-create the error (steps described below). It appears that I'm not understanding the problem.

Using: apachesolr.module,v 1.1.2.32 2008/09/16 11:26:58 robertDouglass Exp $

note: In comment #4 you said "In the meantime I added a check to the latest DRUPAL-5 version that should just avoid the problem by skipping nodes that don't load"

could this be:

static function update_index($namespace) {
....
....
      $node = node_load($row->nid);
      if ($node->nid) {

STEPS that did not re-create the error.

0) Created a few cck fields on Story content type, including a node reference field
1) Noticed that it is not listed with the other fields on admin/settings/apachesolr/index
2) Populated a number of records, referencing nodes on a few
3) Removed a couple of the referenced nodes
4) Used "Delete the Index" button and confirmation checkbox on admin/setting/apachesolr/index
5) Ran cron from Devel menu a few times to get 100% Indexed
6) Recent Log Entries shows: (no errors)

cron 09/17/2008 - 17:57 Cron run completed. admin
Apache Solr 09/17/2008 - 17:57 Adding 0 documents in cleanup admin
cron 09/17/2008 - 17:57 Cron run completed. admin
Apache Solr 09/17/2008 - 17:57 Adding 6 documents in cleanup admin
Apache Solr 09/17/2008 - 17:57 Adding 50 documents admin
cron 09/17/2008 - 17:56 Cron run completed. admin
Apache Solr 09/17/2008 - 17:56 Adding 0 documents in cleanup admin
Apache Solr 09/17/2008 - 17:56 Adding 50 documents admin

Can the authors of this ticket please confirm the problem using the newest check out from the DRUPAL-5 branch?