For those that are not regulars on the mailing list, I am a fairly active
member that has used Elasticsearch for years.
I am leaving my full-time job to focus on other (techie and non-techie)
goals and would love to work on some interesting projects part-time. It can
be either paid assignments or
to the fantastic Elasticsearch team who did not
hesitate to test the fix immediately and replaced it with a better working
solution, since the lzf-compress software is having weaknesses regarding
threadsafety.
Jörg
On Wed, Aug 27, 2014 at 7:01 PM, Ivan Brusic iv...@brusic.com wrote:
Amazing job
,
start_offset : 103,
end_offset : 112,
type : word,
position : 18
} ]
}
So it seems the template is not used?! Any obvious reason/mistakes?
Thx,
Marc
On Thursday, August 28, 2014 6:17:08 PM UTC+2, Ivan Brusic wrote:
Use the Analyze API to view what tokens are being
I used to apply that trick all the time with older versions of
Elasticsearch! Thankfully it has not occurred to me in years.
--
Ivan
On Thu, Aug 28, 2014 at 3:53 PM, Mark Walkom ma...@campaignmonitor.com
wrote:
Yep, the easiest way is to drop the replica and then add it back and see
how
.
Regards,
Mark Walkom
Infrastructure Engineer
Campaign Monitor
email: ma...@campaignmonitor.com
web: www.campaignmonitor.com
On 30 August 2014 05:16, Ivan Brusic i...@brusic.com wrote:
The replica of a shard should never be on the same node as the primary.
Where did you notice this anomaly
appId=cs Times=Me:22/Total:22 (updated
attributes=gps_lng: 183731222/ gps_lat: 289309222/ ) I cannot search for
MyMDB or onMessage; only MyMDB.onMessage will work.
Anymore Ideas?
Cheers,
Marc
On Wednesday, August 27, 2014 9:20:49 AM UTC+2, Ivan Brusic wrote:
Off the top of my head, I
A dismax query will basically rewrite the query as a boolean query. Can you
create your own query where one of the clauses has a negative boost? Still
tricky since you basically need the inverse of cross/best_field where the
field remains the same, but the query changes.
--
Ivan
On Thu, Aug
Also note that the content returned will still contain the stop words. Only
the inverted index will contain the stopword-less content.
--
Ivan
On Thu, Aug 28, 2014 at 11:55 AM, Itamar Syn-Hershko ita...@code972.com
wrote:
What would be the usecase for such a process (removing stop words
:03 GMT-05:00 Ivan Brusic i...@brusic.com:
Also note that the content returned will still contain the stop words.
Only the inverted index will contain the stopword-less content.
--
Ivan
On Thu, Aug 28, 2014 at 11:55 AM, Itamar Syn-Hershko ita...@code972.com
wrote:
What would
the
only way to remove stop words from tokens obtained from a keyword tokenizer?
Are those regular expressions not very performant?
2014-08-28 15:49 GMT-05:00 Ivan Brusic i...@brusic.com:
You mentioned in your original post I'd like to obtain the original
text without stop words
Off the top of my head, I would use a custom analyzer with a whitespace
tokenizer and a word delimiter filter (preserving the original tokens as
well). Perhaps a shingle filter to create bigrams. Or better yet a pattern
tokenizer with spaces and parenthesis.
Cheers,
Ivan
On Tue, Aug 26, 2014
I have a couple of questions regarding function scores. I probably already
know the answers, but just wanted to double-check with the community. I am
still on version 1.1.1, so perhaps things have changed since then.
First question is regarding ordering and efficiency. Currently, my query is
a
Amazing job. Great work.
--
Ivan
On Tue, Aug 26, 2014 at 12:41 PM, joergpra...@gmail.com
joergpra...@gmail.com wrote:
I fixed the issue by setting the safe LZF encoder in LZFCompressor and
opened a pull request
https://github.com/elasticsearch/elasticsearch/pull/7466
Jörg
On Tue,
The function score should not affect relevancy, only the scoring, so the
number of results should not differ. Strange.
Perhaps you do not need to use a function score. With the simple query
string, you can append the boost parameter to the field name:
simple_query_string: {
query: 128,
fields:
The standard use case for a multi-field is when a field needs to be both
analyzed (for searching) and not analyzed (for aggregating/sorting). In
this case, there really is no workaround, so a multi-field is essential.
In the different analyzer case, it gets more complicated. How much can you
get
is not working. If you post some example
documents and mapping, but others might be able to figure it out.
--
Ivan
On Wed, Aug 27, 2014 at 11:07 AM, Ivan Brusic i...@brusic.com wrote:
The function score should not affect relevancy, only the scoring, so the
number of results should not differ
This process is easier (but still not easy) if you pre-process your data on
the client side at indexing time. You can mark your terms with their
respective Parts of Speech using a payload filter:
A few questions:
What version of Elasticsearch are you using?
Are you using the Java client and is it the same version of the cluster?
Did you upgrade recently and was the index built with an older version of
Elasticsearch?
Elasticsearch recently added checksum verification (1.3?), so perhaps
, and with the current transport
request/response cycle, they must poll for new events ...
Jörg
On Thu, Jul 10, 2014 at 6:38 PM, Ivan Brusic iv...@brusic.com wrote:
Jörg, have you actually implemented your own ClusterStateListener? I
never had much success. Tried using that interface or even
allocated 5*10G RAM to
the cluster. Things are looking ok as of now, except that the aggregations
(on strings) are quite slow. May be I would run these aggregations as batch
and cache the outputs in a different type and move on for now.
Thanks
NY
On Fri, Aug 22, 2014 at 10:34 PM, Ivan Brusic
I suspect the issue is the way the query parser works. The query phrase
exampleof bug will be parsed into a query for the tokens exampleof and
bug that are adjacent to each other. The issue is that you do not have
two such tokens, instead you have a token with the value exampleof bug,
which is a
documentation.
Thanks!
On Thursday, August 21, 2014 9:52:16 AM UTC-7, Ivan Brusic wrote:
I suspect the issue is the way the query parser works. The query phrase
exampleof bug will be parsed into a query for the tokens exampleof and
bug that are adjacent to each other. The issue is that you do
Here is the Lucene issue: https://issues.apache.org/jira/browse/LUCENE-2605
--
Ivan
On Thu, Aug 21, 2014 at 10:09 AM, Ivan Brusic i...@brusic.com wrote:
The query string query is a phrase query \exampleof bug\
The term query is looking for a single token exampleof bug
The query parser
, August 21, 2014 10:09:29 AM UTC-7, Ivan Brusic wrote:
The query string query is a phrase query \exampleof bug\
The term query is looking for a single token exampleof bug
The query parser will not use your tokenizer to parse the phrase. It will
tokenize based on whitespace and then apply
.
Thanks for your help.
On Thursday, August 21, 2014 10:42:32 AM UTC-7, Ivan Brusic wrote:
In general, if you are using the keyword tokenizer or non analyzed
fields, then query string queries should probably not be used. Phrase
queries and the keyword tokenizer also do not mix well.
Your
AFAIK, there is no way to achieve such functionality.
The only way I have figured out have similar functionality is to write a
plugin with a cluster state listener and have the plugin reach out to some
external service.
Cheers,
Ivan
On Thu, Aug 21, 2014 at 10:02 AM, 'Sandeep Ramesh Khanzode'
Very strange query indeed. Wildcard search filtered by a match_all. What?!?
It is not Elasticsearch, but perhaps some plugin. Itamar mentioned Kibana,
although you did not mention it in your post. Any other plugins? Marvel?
--
Ivan
On Wed, Aug 20, 2014 at 12:43 PM, Itamar Syn-Hershko
If you only have 3 nodes, I would just stick to the defaults, which is both
master and data.
Having dedicated master (no data) nodes helps because it helps eliminate
OOM pressures since the actual data lives elsewhere. With so few nodes,
every machine should hold a portion of the data. Dedicated
letters? (note that we actually regard 'ij' to be a
single character.) It's not like removing the accents from 'ä', or
transcribing a Cyrillic number into it's Roman equivalent, or am I wrong to
that regard?
Regards,
Matthias
On Tuesday, August 19, 2014 6:37:29 AM UTC+2, Ivan Brusic wrote
At one point Elasticsearch shaded several different libraries for various
reasons, but thankfully this is no longer the case. From what I understand,
the Jetty classes you are referring to are custom classes built for
Elasticsearch that are not packaged with Jetty.
Cheers,
Ivan
On Tue, Aug 19,
at the index
with non html tags if you will?
On Friday, August 8, 2014 12:52:37 PM UTC-4, Ivan Brusic wrote:
The field is derived from the source and not generated from the
tokens.
If we indexed the sentence The quick brown foxes jumped over the
lazy dogs with the english analyzer, the tokens
Char filters are applied before the text is tokenized, and therefore they
are applied before the normal filters are used, which is why they are a
separate class of filter. With Lucene, the order is:
char filters - tokenizer - filters
Have you looked into the ICU analyzer?
Trying using a filter aggregation:
http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/search-aggregations-bucket-filter-aggregation.html
The idea is that the filter is the outer most aggregation and the
aggregation you actually want to filter is the sub-aggregation.
Cheers,
Look into the Lucene query parser, which its the syntax that the query
string query uses. After that, look into the various Lucene contrib modules
that extend the query syntax (span near is one).
I do not think that anyone has implemented a new query parser as an
elasticsearch plugin yet, but I
If the 200kb number is fixed, then the simplest solution would be to store
that content separately in a new field. It does not need to be analyzed,
just stored.
Perhaps highlighters might work. Never used them, so it is just a guess.
Cheers,
Ivan
On Aug 12, 2014 8:17 AM, Dmitriy Bashkalin
You can wrap each individual match query in a constant score query and place
them as clauses in a boolean query.
The guide has an example:
http://www.elasticsearch.org/guide/en/elasticsearch/guide/current/ignoring-tfidf.html#constant-score-query
Cheers,
Ivan
Bernhardt Scherer
The analyzers control how text is parsed/tokenized and how terms are
indexed in the inverted index. The source document remains untouched.
--
Ivan
On Fri, Aug 8, 2014 at 9:24 AM, IronMike sabdall...@gmail.com wrote:
I also used Clint's example and tried to map it to a document and search
not for the
source. I thought I could query the field not the source and look at it
with no html while the source was intact. Did I misunderstand?
On Friday, August 8, 2014 12:36:16 PM UTC-4, Ivan Brusic wrote:
The analyzers control how text is parsed/tokenized and how terms are
indexed
The problem might be with your encoding, not the analyzer. Your content is
in one format and either your output is in another or your viewer
(terminal, browser) is in another. Make sure everything is consistent
(UTF-8 for most people).
Where are you seeing the � character?
--
Ivan
On Thu, Aug
Since version 1.0, there should be fewer binary protocol issues between any
nodes, including the clients, making rolling upgrades doable. Older clients
should be able to interact with newer server nodes, but the inverse is not
always the case.
--
Ivan
On Wed, Aug 6, 2014 at 8:47 AM, Brian
1. Correct.
2. Also correct. The analysis chain only affects how the terms are indexed
and placed in the inverted index. The original document remains as is.
3. Not sure since I have never done highlighting. Highlighting might not
depend on the source since the term positions/offsets are used, but
Perhaps the top hits aggregation can help:
http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/search-aggregations-metrics-top-hits-aggregation.html
--
Ivan
On Wed, Aug 6, 2014 at 11:21 AM, slavag slav...@gmail.com wrote:
Hi,
Need some advise.
I have indexed documents,
: [
*
]
},
size : 1
}
}
}
}
}
}
}
What could be issue with my aggregation ? I'm using ES 1.2.1
Thanks
On Wednesday, August 6, 2014 10:06:40 PM UTC+3, Ivan Brusic wrote
I am still fully in the nothing but E stack! Is anyone else using
Elasticsearch for ... search? :)
--
Ivan
On Tue, Aug 5, 2014 at 10:50 AM, Brian brian.from...@gmail.com wrote:
Using the most recent release (1.2.2) of Curator, I noticed that the
documentation says --logfile while curator
Are you applying a custom routing to your documents?
--
Ivan
On Tue, Aug 5, 2014 at 2:33 AM, Warat Wongmaneekit canopyb...@gmail.com
wrote:
Now my cluster is not rebalance the data. How can I rebalance it please
see the summary below.
There should be no need to run a master and a data node on each machine.
Only two masters is not enough to reliably form a consensus and you are
only taking away processing power from the data node.
--
Ivan
On Tue, Aug 5, 2014 at 5:30 AM, Ankit Mittal ankit.lnc...@gmail.com wrote:
Hi All,
Are you refreshing the index after inserting the test documents? I could be
simply a matter of timing.
--
Ivan
On Sun, Aug 3, 2014 at 8:22 AM, John D. Ament john.d.am...@gmail.com
wrote:
Hi
So after running a few rounds of local automated tests, I've noticed that
sometimes I get the wrong
The only way to achieve the result you are seeking is to use parent/child
documents:
http://www.elasticsearch.org/blog/managing-relations-inside-elasticsearch/
http://www.spacevatican.org/2012/6/3/fun-with-elasticsearch-s-children-and-nested-documents/
Cheers,
Ivan
On Mon, Aug 4, 2014 at 9:10
You should switch to using bulk indexing instead of indexing an individual
documents. Also, considering switching off the refresh interval (set it to
-1) for the duration of your bulk indexing.
Cheers,
Ivan
On Mon, Aug 4, 2014 at 3:08 AM, Dennis de Boer datdeb...@gmail.com wrote:
Sure,
Javadocs also available at
http://jenkins.elasticsearch.org/job/Elasticsearch%20Master%20Branch%20Javadoc/Elasticsearch_API_Documentation/
http://javadoc.kyubu.de/elasticsearch/ (unofficial)
--
Ivan
On Mon, Aug 4, 2014 at 5:28 AM, Bernd Fehling bernd.fehl...@gmail.com
wrote:
Thanks a lot.
are fixed on
processor size, is there any chance of automatic updation on these?
Thanks again for reporting this.
On Thursday, 31 July 2014 22:41:19 UTC+5:30, Ivan Brusic wrote:
The thread pool will reject any search requests when there are 1000
actions already queued.
http
With inner dccuments, you cannot query two different attributes for the
same inner object. Internally, Elasticsearch/Lucene will flatten all the
inner properties into one array. You would need to switch to nested
documents or even parent/child documents in order to query on a single
deeper
Once I hit 3 open pull requests, I stopped submitting them. :) I am down to
only 1 open (from last year), but my confidence is still somewhat low. Most
of my changes have been small, mainly because I am hesitant about
dedicating more time for something that might be ignored.
I need to simply
I just deleted a couple of non-related job postings and banned the poster.
Going forward, is there a consensus among the community about whether or
not job postings should be allowed? I do not mind postings that come
directly from companies, especially those whose existing developers are
already
That scenario should not happen since FileVisitOption.FOLLOW_LINKS is
enabled.
https://github.com/elasticsearch/elasticsearch/blob/fe86c8bc88a321bf587dd8eb4df52aaed9ed2156/src/main/java/org/elasticsearch/common/logging/log4j/LogConfigurator.java#L107
Seems like a bug somewhere.
--
Ivan
On
All types eventually belong to the same Lucene index and Lucene cannot
handle different types for the same field name. Avoid using the same name
across types if the field type is different.
http://www.elasticsearch.org/guide/en/elasticsearch/guide/current/mapping.html#_avoiding_type_gotchas
--
Tuning stop words can be as long of a process as you want it to be. Saving
your queries/results and doing some search analytics can help you fine tune
the stop words. In general, the default stop words list is very good for
English, but Twitterspeak is not really English. :)
You can look at all
The thread pool will reject any search requests when there are 1000 actions
already queued.
http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-threadpool.html
Do you have this many search requests at one time? Do you have warmers
and/or percolators running since you
Look into the curator, which should help:
https://github.com/elasticsearch/curator
If you have just a single development instance, perhaps Marvel is an
overkill. Do you need historical metrics? If not, just use some other
plugin such as head/bigdesk/hq.
Cheers,
Ivan
On Thu, Jul 31, 2014 at
You should as many nodes as possible. If you enable client.transport.sniff,
then the transport client will ask the nodes it does connect to about the
other nodes in the cluster, which means you can potentially only need to
specific a single node (not ideal in case that node is down).
--
Ivan
The logging.xml file will only control which logging statements get
outputed, not the amount of information it may contain.
The log line in question does not have the source ip, which is long gone by
the time the service gets the request.
, Andrew Davidoff david...@qedmf.net wrote:
On Tuesday, July 29, 2014 3:27:13 PM UTC-4, Ivan Brusic wrote:
Have you changed your gateway settings? http://www.
elasticsearch.org/guide/en/elasticsearch/reference/
current/modules-gateway.html#recover-after
It still remains a bit of black magic to me
It will depend on your merge settings and your shard size. Not sure why
not, but I do not recall what the default settings are.
--
Ivan
On Mon, Jul 28, 2014 at 8:52 PM, smonasco smona...@gmail.com wrote:
Do you happen to know if optimize will create a segment larger than 5 gigs?
--
You
The _all field has its own analyzer, so the analyzer that is defined on the
createdBy field is not applied.
I have never tried, but I believe the best solution is to use copy-to to
a custom field:
on the client side. My application
knows which fields are analyzed and which are not and creates queries
accordingly.
--
Ivan
On Tue, Jul 29, 2014 at 10:52 AM, Ivan Brusic i...@brusic.com wrote:
The _all field has its own analyzer, so the analyzer that is defined on
the createdBy field
Would it be possible to script the sort order of aggregations? Let me
explain with a contrived example:
{
aggs: {
agg1: {
terms: {
field: somefield,
order: {
avg1: desc
},
size : 100
},
aggs: {
You can have an alias point to multiple indices. With time-series data,
this should be a problem since you will not have overlap between the
different indices. But I think you are correct in that there is no atomic
way to accomplish all three.
I only search against aliases, so having an index
Typo: With time-series data, this should NOT be a problem
On Tue, Jul 29, 2014 at 11:57 AM, Ivan Brusic i...@brusic.com wrote:
You can have an alias point to multiple indices. With time-series data,
this should be a problem since you will not have overlap between the
different indices. But I
Have you changed your gateway settings?
http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-gateway.html#recover-after
It still remains a bit of black magic to me. Sometimes it works, sometimes
it does not.
Cheers,
Ivan
On Mon, Jul 28, 2014 at 1:52 PM, Andrew
, Adrien Grand
adrien.gr...@elasticsearch.com wrote:
Hi Ivan,
This is not supported but there is an open issue that aims at making this
kind of things possible
https://github.com/elasticsearch/elasticsearch/issues/6917
On Tue, Jul 29, 2014 at 8:50 PM, Ivan Brusic i...@brusic.com wrote
The behavior is applied at the creation of the index within the mapping,
not during the prepareIndex call. The example you provided is part of the
mapping and not part of the document that gets indexed. If you want to
override which field will be used as the _id field, you would need to
change
There was a bug in Lucene which caused problems with Elasticsearch 1.3.0.
You might already know this, but 1.3.1 was released today to fix this issue:
http://www.elasticsearch.org/blog/elasticsearch-1-3-1-released/
The issue should only affect older versions. Your version is newer, but the
error
Answers inline.
On Fri, Jul 25, 2014 at 3:06 AM, CB chen.be...@gmail.com wrote:
thanks for the answers, here are my thoughts:
1. If using pure REST client - Using a Load Balancer will make sure that
the endpoint address goes to any of the live nodes (round robin) so that
if one of those
or configuration to
completely skip the score evaluation process.
Thank you, Ivan!
Ivan Brusic於 2014年7月25日星期五UTC+8上午1時03分53秒寫道：
I am not sure if there is a cleaner way to bypassing score, but if you
explicitly sort against another value that is not the score, then by
default scoring
I am not sure if there is a cleaner way to bypassing score, but if you
explicitly sort against another value that is not the score, then by
default scoring will not occur.
Perhaps if you trace the code for sort, you can find a setting that
disables scoring in general.
--
Ivan
On Wed, Jul 23,
Thanks for chiming in Renaud. Hopefully I will have a chance to test out
the plugin soon. My use case for nested documents is fairly simple.
--
Ivan
On Thu, Jul 24, 2014 at 4:00 AM, ren...@sindicetech.com wrote:
Hi Brian,
Our apologies for the issues with the web site, we had some problems
1. You can retrieve the term position, offset and payload using function
score scripts:
http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-advanced-scripting.html
2: There are a couple of proposed solutions that would store the data in
another index that is joined with
In the end, all the documents end up in the same Lucene index, and while
Lucene is schema-less, all similarly named fields must be the same type.
Types are useful in Elasticsearch to separate different type
configurations, but will fail on similarly named fields.
There is some work being done to
Has anyone else seen this plugin? http://siren.solutions/siren/overview/
There was some discussion between one of the developers and Jorg a while
back, so I guess this is the outcome. Have not tried it yet, but I will
give it a shot this weekend. I am hoping that it can fix a longstanding
issue
You can wrap any query with a query filter:
http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/query-dsl-query-filter.html
--
Ivan
On Wed, Jul 23, 2014 at 1:52 PM, IronMike sabdall...@gmail.com wrote:
How can I exclude exact phrases with a filter?
Lets say I want to
Your issue is casing. You are only applying the synonym filter, which by
default does not lowercase terms. You can either set ignore_case to true
for the synonym filter or apply a lower case filter before the synonym. I
prefer to use the latter approach since I prefer to have all my analyzed
}
}
}
}
}'
On Tuesday, July 22, 2014 11:56:40 AM UTC-5, Ivan Brusic wrote:
Your issue is casing. You are only applying the synonym filter, which by
default does not lowercase terms. You can either set ignore_case to true
for the synonym filter or apply a lower case filter before
-tokenstreams-are-actually.html
http://nolanlawson.com/2012/10/31/better-synonym-handling-in-solr/
--
Ivan
On Tue, Jul 22, 2014 at 11:03 AM, Ivan Brusic i...@brusic.com wrote:
A couple of reasons. The biggest issue is multi word synonyms since the
query parser will tokenize the query before
The post filter is not a addition in 0.90.8, just a renaming of a field
that was ambiguous:
https://github.com/elasticsearch/elasticsearch/issues/4119
Pre filters are simply filtered queries. In most cases, you want to use the
pre filters. Queries are expensive in Lucene since you have to score
Your mapping does not seem correct. Can you post the output of the get
mapping API instead? It appears that the region field might be a geo type
instead. Analyzed fields should have position data enabled by default.
--
Ivan
On Tue, Jul 22, 2014 at 8:46 PM, xu piao xupia...@gmail.com wrote:
i
Recovery is throttled since version 0.90.1
http://www.elasticsearch.org/guide/en/elasticsearch/reference/0.90/index-modules-store.html#store-throttling
Increase indices.store.throttle.max_bytes_per_sec to a level that is
suitable for your environment. Since IO should be the main bottleneck, the
Most users do not set the direct memory setting. mlockall is set, but does
the server allow it? You would see an error on startup if it didn't. Did
you change the vm swapiness on the server?
--
Ivan
On Wed, Jul 16, 2014 at 2:40 AM, Pedro Jerónimo pedropregue...@gmail.com
wrote:
*Java: *java
By default, string fields are analyzed using the standard analyzer, which
will tokenize and lowercase the input (I believe stop words are now NOT
removed). A term query does not analyze the query, so it only works on non
analyzed fields (or fields that use a keyword tokenizer). A term query for
what the problem is for
that query.
On Wed, Jul 16, 2014 at 10:27 AM, Ivan Brusic i...@brusic.com wrote:
By default, string fields are analyzed using the standard analyzer, which
will tokenize and lowercase the input (I believe stop words are now NOT
removed). A term query does not analyze
behaving as if it has been
analyzed.
On Wed, Jul 16, 2014 at 11:00 AM, Ivan Brusic i...@brusic.com wrote:
I would verify that the field is in fact non_analyzed and that your data
is
indexed in the way you think it is. Use the analyze API to analyze the
term.
Make sure you use the last
As predicted, your actual mapping does not match your perceived mapping.
Something is not matching up. Perhaps the mapping is for a different index
or type. Best way is to share your mapping and perhaps how you created your
index as indicated at http://www.elasticsearch.org/help
--
Ivan
On
Read more about it here:
http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/index-modules-store.html#store-throttling
Previously it was unbounded, but now the default is 20mb, which I found to
be extremely low. Also, prior to 1.2, there was a Lucene bug which made
throttling
Direct memory is off heap memory. Are elasticsearch and logstash the only
processes on those servers? Did you set an explicit direct memory value?
--
Ivan
On Jul 15, 2014 3:46 PM, Mark Walkom ma...@campaignmonitor.com wrote:
How much data do you have in ES, index count and total size of all
Since the script is executed against lots of matched documents, perhaps
converting it into a native Java script (not Javascript) would provide a
performance boost.
Note that using fields in scripts will force their values to be loaded into
the cache.
--
Ivan
On Sun, Jul 13, 2014 at 8:54 AM,
There are a few settings where the full named is not specified in the code,
but is relative to the module it is in. Does your grep code account for
these settings?
A repo with pull requests might be too much for the maintainer, but a wiki
would work well.
Great job,
Ivan
On Mon, Jul 14, 2014
Jörg is correct. In general, it would be a bad idea to change the
similarity during runtime, but there are cases were it would be acceptable
and the system should allow for those cases:
https://github.com/elasticsearch/elasticsearch/issues/4403
--
Ivan
On Mon, Jul 14, 2014 at 12:00 AM, Jörg
This technically sounds like a Kibana question, so you might have better
luck with the Logstash mailing list.
Can't you simply prepend the field name in the query instead of relying on
the default field? You can also change field names in Logstash.
Another option is the copy-to-field. Similar to
First of all, there is no version 0.26. I am assuming you meant 0.20.6.
Either way, any upgrade from prior of 1.0 to 1.x will require a full
cluster restart.
1. No clue
2. Many settings like omit_norms were deprecated, but are still support. I
think that omit_tf has been changed.
3. I would
Read the recent comments regarding with the recent spam:
https://groups.google.com/d/msg/elasticsearch/byATcjKgdYE/_Neoiof4fKIJ
This new spam account has been banned.
--
Ivan
On Sat, Jul 12, 2014 at 2:50 PM, Warner Onstine warn...@gmail.com wrote:
Could we please turn on first post filters
Besides stop works, you can use a bool query one clause is the match all,
and the other clause is must not with the terms in question.
Something like:
{
query: {
bool: {
must: [
{
match_all: {}
}
],
must_not: [
101 - 200 of 513 matches
Mail list logo
