Ok Karl, thanks for the tip and the quick response, we will do this and come back with the result.

On Wed, Nov 6, 2013 at 9:28 PM, Karl Wright <daddywri@gmail.com> wrote:
Hi Ronny,

One minor thing: you should need to set throttling to 2 ONLY for the Windows repository connection, not for AD or Solr.

As for how to debug this issue, first off you should be looking in the manifoldcf.log file (or the equivalent).  You should see WARN messages from the shared file connector under most conditions when there's a service interruption.  You would probably see "Read timed out" warnings if you looked there, since that is what aborted the job run, along with a stack trace.  However, that's not going to add much information to the analysis at this point.

What might be valuable is to determine whether the problem is happening on the Windows side or on the Solr side.  At this point I can't tell.  You could, however, create a null output connection, and create  a similar job the sends its output there, and see if it completes.  Can you do this and get back to me?


On Wed, Nov 6, 2013 at 3:17 PM, Ronny Heylen <securaqbereusr@gmail.com> wrote:
We use Manifoldcf 1.3 and Solr 4.4 to index a shared network drive with several hundred thousands documents.
Doing only one manifoldcf job to index all the drive was always giving some kind of error, therefore to better understand where the problem can be, we made one job to index all *.doc*, another one for *.xls*, another one for *.pdf ...
Using the help from the list (thanks!) we set the size limit to 100MB and all jobs succeeds (great) except the one for *.pptx
The message is
Error: Repeated service interruptions - failure processing document: Read timed out
We don't find any error in the log we have searched: solr.log, ...
Based on some indications found on Internet, we have set the Throttling max connections setting to 2 (instead of 10) in 3 places:
output connection to SOLR
authority connection to the Active Directory
repository connection to the windows file share
But the problem stays the same.
We have tried on another machine with SOLR 4.5 and Manifoldcf 1.4, same problem.
We can let run the job for all *.PDF, or all *.DOC*, or all *.XLS* without problem, but the same message comes always for *.PPTX.
The last time the job stops with the message, it displays (not the same numbers for each run as the windows drive is changing) 56311 documents, with 17466 busy and 38847 processed.
As we don't find anything in the log (but probably we don't look at the correct place), we don't know what to do.
Thanks for your help,
Ronny and Frédéric