dovecot - Oct 2011 - Rare error with concurrent users searching with fts-solr

Hi all,

If I start two o more connections with diferent users that make a
search text to solr I get this message many times:

Error: fts_solr: Query didn't return uid

If I repeat the test about 2000 times, with only one user concurrent,
with 60 diferent users, without concurrent, I get between 0.

If I repeat the test about 2000 times, with 2 users concurrent, I get
between 3 or 10 errors.

If I repeat the test about 2000 times, with 4 users concurrent, I get
between 10-20 errors.


After see this error, I set a tcpdump to try to watch the SOLR XML and
GET, I can watch that Solr send a xml like this:

<response></response>

with 200 as http status

When the normal response is like this

<response>
<lst name="responseHeader">
<int name="status">0</int>
<int name="QTime">1</int>
<lst name="params">
     <str name="fl">uid,score</str>
     <str name="sort">uid asc</str>
     <str name="q">BLABLABLA</str>
     <str name="fq">+user:"ANYUSER" +uidv:1208469916
+box:"ANYFOLDER"
-ns:[* TO *]</str>
     <str name="rows">1</str>
      </lst>
      </lst>
      <result name="response" numFound="0"
start="0" maxScore="0.0"/>
</response>


At first, I think that the problems it's in Solr, but after get all
GET from tcpdump, I make a python script with urllib2 and send it all
to Solr with the same concurrency and I can't get any error or "empty
response"

Have someone made a intensive concurrency tests over Dovecot + Solr ?


-- 
Antonio P?rez-Aranda Alcaide
aperezaranda at yaco.es

Yaco Sistemas S.L.
http://www.yaco.es/
C/ Rioja 5, 41001 Sevilla
Tel?fono +34 954 50 00 57
Fax ? ? ?+34 954 50 09 29

On Thu, 2011-10-13 at 10:31 +0200, Antonio Perez-Aranda
wrote:
> After see this error, I set a tcpdump to try to watch the SOLR XML and
> GET, I can watch that Solr send a xml like this:
> 
> <response></response>
> 
> with 200 as http status
> 
> When the normal response is like this
> 
> <response>
> <lst name="responseHeader">
..

Check what kind of commands Dovecot sends. Is this maybe a reply to a
somewhat different (broken?) command than others?
> At first, I think that the problems it's in Solr, but after get all
> GET from tcpdump, I make a python script with urllib2 and send it all
> to Solr with the same concurrency and I can't get any error or
"empty
> response"
You mean send the exact same commands as Dovecot sent when this failed?

It's anyway difficult for me to guess how this could be Dovecot's fault,
if the command that gives the broken response normally works.