just wanted to be clear on how to interpret the # shard failures during a
search, in particular a Scan type request in case it makes a difference.
Does one interpret that a search response with a # shard failures > 1 is a
'best effort' search response, and perhaps not fully accurate, or is it
simply reporting failures ES has managed to recover from and return a
full/complete response ?
I guess for my case, if I'm really needing the response to be accurate, that
any shard failure count > 0 should require retrying the request (with
bailout/backoff/whatever) built in?
When you get shard failures, it means that the search failed to execute on
the respective shards (on all those shard replicas, as they are
automatically retried) that are reported. The result of the search will
include hits from the rest (non failed) shards.
just wanted to be clear on how to interpret the # shard failures during a
search, in particular a Scan type request in case it makes a difference.
Does one interpret that a search response with a # shard failures > 1 is a
'best effort' search response, and perhaps not fully accurate, or is it
simply reporting failures ES has managed to recover from and return a
full/complete response ?
I guess for my case, if I'm really needing the response to be accurate,
that any shard failure count > 0 should require retrying the request (with
bailout/backoff/whatever) built in?
When you get shard failures, it means that the search failed to execute
on the respective shards (on all those shard replicas, as they are
automatically retried) that are reported. The result of the search will
include hits from the rest (non failed) shards.
ok, so any shard failure is effectively a compromised search, you can't
trust the results. thanks.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.