Apache
/
lucene
mirror of https://github.com/apache/lucene.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Ref Guide: fix note for atomic updates after SOLR-9530

This commit is contained in:
Cassandra Targett 2017-05-23 11:05:18 -05:00
parent 73aa53b063
commit 31e02e93a5
1 changed files with 10 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
solr/solr-ref-guide/src/update-request-processors.adoc
View File

@ -90,7 +90,7 @@ Do not forget to add `RunUpdateProcessorFactory` at the end of any chains you de
Update request processors can also be configured independent of a chain in `solrconfig.xml`.
.updateProcessor
.updateProcessor Configuration
[source,xml]
----
<updateProcessor class="solr.processor.SignatureUpdateProcessorFactory" name="signature">
@ -105,7 +105,7 @@ Update request processors can also be configured independent of a chain in `solr
In this case, an instance of `SignatureUpdateProcessorFactory` is configured with the name "signature" and a `RemoveBlankFieldUpdateProcessorFactory` is defined with the name "remove_blanks". Once the above has been specified in `solrconfig.xml`, we can be refer to them in update request processor chains in `solrconfig.xml` as follows:
.updateRequestProcessorChains and updateProcessors
.updateRequestProcessorChain Configuration
[source,xml]
----
<updateProcessorChain name="custom" processor="remove_blanks,signature">
@ -135,7 +135,7 @@ In summary:
In the previous section, we saw that the `updateRequestProcessorChain` was configured with `processor="remove_blanks, signature"`. This means that such processors are of the #1 kind and are run only on the forwarding nodes. Similarly, we can configure them as the #2 kind by specifying with the attribute "post-processor" as follows:
.post-processors
.post-processor Configuration
[source,xml]
----
<updateProcessorChain name="custom" processor="signature" post-processor="remove_blanks">
@ -145,19 +145,19 @@ In the previous section, we saw that the `updateRequestProcessorChain` was confi
However executing a processor only on the forwarding nodes is a great way of distributing an expensive computation such as de-duplication across a SolrCloud cluster by sending requests randomly via a load balancer. Otherwise the expensive computation is repeated on both the leader and replica nodes.
// TODO 6.6 I think this can be removed after SOLR-9530 -CT
.Pre-processors and Atomic Updates
[WARNING]
====
Because `DistributedUpdateProcessor` is responsible for processing <<updating-parts-of-documents.adoc#updating-parts-of-documents,Atomic Updates>> into full documents on the leader node, this means that pre-processors which are executed only on the forwarding nodes can only operate on the partial document. If you have a processor which must process a full document then the only choice is to specify it as a post-processor.
====
.Custom update chain post-processors may never be invoked on a recovering replica
[WARNING]
====
While a replica is in <<read-and-write-side-fault-tolerance.adoc#ReadandWriteSideFaultTolerance-WriteSideFaultTolerance,recovery>>, inbound update requests are buffered to the transaction log. After recovery has completed successfully, those buffered update requests are replayed. As of this writing, however, custom update chain post-processors are never invoked for buffered update requests. See https://issues.apache.org/jira/browse/SOLR-8030[SOLR-8030]. To work around this problem until SOLR-8030 has been fixed, *avoid specifying post-processors in custom update chains*.
====
=== Atomic Updates
If the `AtomicUpdateProcessorFactory` is in the update chain before the `DistributedUpdateProcessor`, the incoming document to the chain will be a partial document.
Because `DistributedUpdateProcessor` is responsible for processing <<updating-parts-of-documents.adoc#updating-parts-of-documents,Atomic Updates>> into full documents on the leader node, this means that pre-processors which are executed only on the forwarding nodes can only operate on the partial document. If you have a processor which must process a full document then the only choice is to specify it as a post-processor.
[[UpdateRequestProcessors-UsingCustomChains]]
== Using Custom Chains