[
https://issues.apache.org/jira/browse/SOLR-6666?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14255107#comment-14255107
]
Liram Vardi commented on SOLR-6666:
-----------------------------------
Hi Erick,
Thanks you for the comprehensive review for this post.
Indeed, my patch causes a loading failure and also to
TestFieldCollectionResource to fail, when we are using the “fail_dynamic”
example.
(TestManagedSotpFilterFactory and TestManagedSynonymFileFactory failed on my
environment regardless the patch.)
Although, based on the following wiki, I was sure that this example is invalid.
https://cwiki.apache.org/confluence/display/solr/Copying+Fields
However, based on your explanation I tried to find a combine solution which
satisfies those two cases, as you said.
The case that I am trying to solve is the case that source is not explicit
(which means that does not have a field definition and it is only instantiated
by matching a dynamic field – the second case that you described on your
response).
So, let make things a bit more ordered.
Let us assume three possible types for each copyfield, source and destination:
1) Explicit – the field has explicit “field” definition.
2) Glob – The field contains an asterisks on its copyField definition and
it matches to one (or more) of the fields definitions (dynamic or static).
3) Reference – the copy field references to some dynamic field, but it is
without any asterisks.
Each copyfield’s source and destination belongs to one of those types.
When Solr reads the schema, it divides the copy fields eventually to two
groups: fixedCopyFields and to dynamicCopyFields.
As I explained before, the “fixedCopyFields” is much less expensive than the
“dynamicCopyFields”.
Now, let define the following decision table:
|Case||Source||Destination||Decision||
|1|Explicit|Explicit|fixedCopyFields|
|2|Explicit|Glob|Error!|
|3|Explicit|Reference|dynamicCopyFields|
|4|Glob|Explicit|dynamicCopyFields|
|5|Glob|Glob|dynamicCopyFields|
|6|Glob|Reference|dynamicCopyFields|
|*7*|*Reference*|*Explicit*|*fixedCopyFields*|
|8|Reference|Glob|dynamicCopyFields|
|9|Reference|Reference|dynamicCopyFields|
As you can see, until today only for case “1” (source and destination are
explicit), Solr put those copy fields on the “static” hash.
On the next version of patch SOLR-6666, I did a refectory on the “if” statement
which divides those copyfields.
At the previous version of the patch, the code throw exception on case 8 (.i.e
fail_dynamic example).
Now after the refectory, case “8” is legal again and case “7”, which is the one
that I am trying to solve, sends those copyfields to the “fixedCopyFields” map.
Another open question is if cases 3 and 9 need also to stay as
“DynamicCopyFields” or can we make the update more efficient by moving those
also to the “static” map… But currently the patch does not change this.
The second version of the patch is attached.
Thanks!
> Dynamic copy fields are considering all dynamic fields, causing a significant
> performance impact on indexing documents
> ----------------------------------------------------------------------------------------------------------------------
>
> Key: SOLR-6666
> URL: https://issues.apache.org/jira/browse/SOLR-6666
> Project: Solr
> Issue Type: Improvement
> Components: Schema and Analysis, update
> Environment: Linux, Solr 4.8, Schema with 70 fields and more than 500
> specific CopyFields for dynamic fields, but without wildcards (the fields are
> dynamic, the copy directive is not)
> Reporter: Liram Vardi
> Assignee: Erick Erickson
> Attachments: SOLR-6666.patch
>
>
> Result:
> After applying a fix for this issue, tests which we conducted show more than
> 40 percent improvement on our insertion performance.
> Explanation:
> Using JVM profiler, we found a CPU "bottleneck" during Solr indexing process.
> This bottleneck can be found at org.apache.solr.schema.IndexSchema, in the
> following method, "getCopyFieldsList()":
> {code:title=getCopyFieldsList() |borderStyle=solid}
> final List<CopyField> result = new ArrayList<>();
> for (DynamicCopy dynamicCopy : dynamicCopyFields) {
> if (dynamicCopy.matches(sourceField)) {
> result.add(new CopyField(getField(sourceField),
> dynamicCopy.getTargetField(sourceField), dynamicCopy.maxChars));
> }
> }
> List<CopyField> fixedCopyFields = copyFieldsMap.get(sourceField);
> if (null != fixedCopyFields) {
> result.addAll(fixedCopyFields);
> }
> {code}
> This function tries to find for an input source field all its copyFields (All
> its destinations which Solr need to move this field).
> As you can probably note, the first part of the procedure is the procedure
> most “expensive” step (takes O( n ) time while N is the size of the
> "dynamicCopyFields" group).
> The next part is just a simple "hash" extraction, which takes O(1) time.
> Our schema contains over then 500 copyFields but only 70 of then are
> "indexed" fields.
> We also have one dynamic field with a wildcard ( * ), which "catches" the
> rest of the document fields.
> As you can conclude, we have more than 400 copyFields that are based on this
> dynamicField but all, except one, are fixed (i.e. does not contain any
> wildcard).
> From some reason, the copyFields registration procedure defines those 400
> fields as "DynamicCopyField " and then store them in the “dynamicCopyFields”
> array,
> This step makes getCopyFieldsList() very expensive (in CPU terms) without any
> justification: All of those 400 copyFields are not glob and therefore do not
> need any complex pattern matching to the input field. They all can be store
> at the "fixedCopyFields".
> Only copyFields with asterisks need this "special" treatment and they are
> (especially on our case) pretty rare.
> Therefore, we created a patch which fix this problem by changing the
> registerCopyField() procedure.
> Test which we conducted show that there is no change in the Indexing results.
> Moreover, the fix still successfully passes the class unit tests (i.e.
> IndexSchemaTest.java).
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
