Thanks for the help!
Yes, we are doing a commit following the update. We will try
IndexWriter.setInfoStream
Below are our the environments we are testing on:
Ubuntu Hardy, Kernel 2.6.16-xenU i386
Amazon EC2, US East Region
Embedded Jetty
Java 1.6.0_16
Solr 1.4
Server B
Ubuntu Hardy, Kernel
Yes, more details would be great...
Is this easily repeated?
The exists?=false is particularly spooky.
It means, somehow, a new segment was being flushed, containing 1285
docs, but then after closing the doc stores, the stored fields index
file (_X.fdx) had been deleted.
Can you turn on
We are using Solr 1.4 in a multi-core setup with replication.
Whenever we write to the master we get the following exception:
java.lang.RuntimeException: after flush: fdx size mismatch: 1285 docs vs 0
length in bytes of _gqg.fdx file exists?=false
at
We need more information. How big is the index in disk space? How many
documents? How many fields? What's the schema? What OS? What Java
version?
Do you run this on a local hard disk or is it over an NFS mount?
Does this software commit before shutting down?
If you run with asserts on do you