So far we've indexed about 46 million documents, but over the weekend,
these errors started coming up. I would expect that if there was a
basic issue, it would have started right away? We ran a test cluster
with just a few shards/replicas prior and didn't see any issues using
the same indexing code, but we're running a lot more indexers
simultaneously with the larger cluster; perhaps we're just overloading
HDFS? The same nodes that run Solr also run HDFS datanodes, but they
are pretty beefy machines; we're not swapping.
As Shawn pointed out, I will be checking the HDFS version (we're using
Cloudera CDH 5.10.2), and the HDFS logs.
On 7/17/2017 10:16 AM, Susheel Kumar wrote: