* Lower fielddata circuit breaker default limit
Lower fielddata circuit breaker default limit from 60% to 40% as we have
moved to doc_values for most of the cases.
* merge master in
* update tests
* update docs
Bulk requests comprise many individual actions, and the responses for each
action comes back in the same order (see e.g. `DocumentActionsIT#testBulk()`).
However the docs do not seem to explicitly state this vital fact. This commit
addresses that omission.
The following updates were made:
- Add a new untyped endpoint `{index}/_explain/{id}`.
- Add deprecation warnings to Rest*Action, plus tests in Rest*ActionTests.
- For each REST yml test, make sure there is one version without types, and another legacy version that retains types (called *_with_types.yml).
- Deprecate relevant methods on the Java HLRC requests/ responses.
- Update documentation (for both the REST API and Java HLRC).
This commit gets rid of the 'NONE' and 'INFO' severity levels for
deprecation issues.
'NONE' is unused and does not make much sense as a severity level.
'INFO' can be separated into two categories: Either 1) we can
definitively tell there will be a problem with the cluster/node/index
configuration that can be resolved prior to upgrade, in which case
the issue should be a WARNING, or 2) we can't, because any issues would
be at the application level, for which the user should review the
deprecation logs and/or response headers.
In real deployments it is important that clusters are properly configured to
avoid accidentally forming multiple independent clusters at cluster
bootstrapping time. However we also expect to be able to unpack Elasticsearch
and start up one or more nodes without any up-front configuration, and have
them do their best to find each other and form a cluster after a few seconds.
This change adds a delayed automatic bootstrapping process to nodes that start
up with no relevant settings set to support the desired out-of-the-box
experience without compromising safety in properly-configured deployments.
* Add deprecation warnings to `Rest*TermVectorsAction`, plus tests in `Rest*TermVectorsActionTests`.
* Deprecate relevant methods on the Java HLRC requests/ responses.
* Update documentation (for both the REST API and Java HLRC).
* For each REST yml test, create one version without types, and another legacy version that retains types (called *_with_types.yml).
This commit changes the format of the `hits.total` in the search response to be an object with
a `value` and a `relation`. The `value` indicates the number of hits that match the query and the
`relation` indicates whether the number is accurate (in which case the relation is equals to `eq`)
or a lower bound of the total (in which case it is equals to `gte`).
This change also adds a parameter called `rest_total_hits_as_int` that can be used in the
search APIs to opt out from this change (retrieve the total hits as a number in the rest response).
Note that currently all search responses are accurate (`track_total_hits: true`) or they don't contain
`hits.total` (`track_total_hits: true`). We'll add a way to get a lower bound of the total hits in a
follow up (to allow numbers to be passed to `track_total_hits`).
Relates #33028
The current response format is:
```
{
"pattern1": {
...
},
"pattern2": {
...
}
}
```
The new format is:
```
{
"patterns": [
{
"name": "pattern1",
"pattern": {
...
}
},
{
"name": "pattern2",
"pattern": {
...
}
}
]
}
```
This format is more structured and more friendly for parsing and generating specs.
This is a breaking change, but it is better to do this now while ccr
is still a beta feature than later.
Follow up from #36049
This change adds a soft limit to open scroll contexts that can be controlled with the dynamic cluster setting `search.max_open_scroll_context` (defaults to 500).
When building a query Lucene distinguishes two cases, queries that require to produce a score and queries that only need to match. We cloned this mechanism in the QueryBuilders in order to be able to produce different queries based on whether they need to produce a score or not. However the only case in es that require this distinction is the BoolQueryBuilder that sets a different minimum_should_match when a `bool` query is built in a filter context..
This behavior doesn't seem right because it makes the matching of `should` clauses different when the score is not required.
Closes#35293
The new limit on the number of open shards in a cluster may be
interpreted by users as a sizing recommendation, but it is not. This
clarifies in the documentation that this is a safety limit, not a
recommendation.
This commit documents how Index Lifecycle Management
interacts with snapshot/restore, and documents a workaround
for situations in which ILM should not immediately resume
managing an index after it is restored.
A number of tokenfilters can produce multiple tokens at the same position. This
is a problem when using token chains to parse synonym files, as the SynonymMap
requires that there are no stacked tokens in its input.
This commit ensures that when used to parse synonyms, these tokenfilters either produce
a single version of their input token, or that they throw an error when mappings are
generated. In indexes created in elasticsearch 6.x deprecation warnings are emitted in place
of the error.
* asciifolding and cjk_bigram produce only the folded or bigrammed token
* decompounders, synonyms and keyword_repeat are skipped
* n-grams, word-delimiter-filter, multiplexer, fingerprint and phonetic throw errors
Fixes#34298
Add a short extra sentence that explains that a missing query part in a search
request containing a "suggest" section will mean only suggestions are returned.
Closes#31640
Move classes under the same package to avoid internal classes being
exposed to the outside. Remove public visibility outside 3 classes:
EsDriver, EsDataSource and EsTypes.
The driver only has one package, namely org.elasticsearch.xpack.sql.jdbc
Use Es prefix for classes to ease name conflict and indicate their
destination
Fix#35437
* [Rollup] Add more diagnostic stats to job
To help debug future performance issues, this adds the
min/max/avg/count/total latencies (in milliseconds) for search
and bulk phase. This latency is the total service time including
transfer between nodes, not just the `took` time.
It also adds the count of search/bulk failures encountered during
runtime. This information is also in the log, but a runtime counter
will help expose problems faster
* review cleanup
* Remove dead ParseFields
`ScriptDocValues#getValues` was added for backwards compatibility but no
longer needed. Scripts using the syntax `doc['foo'].values` when
`doc['foo']` is a list should be using `doc['foo']` instead.
Closes#22919
MultiSearchRequests issues through `_msearch` now validate all keys
in the metadata section. Previously unknown keys were ignored
while now an exception is thrown.
Closes#35869
This commit adds documentation for authorization_realms
setting for the Kerberos realm and also corrects a typo in
existing documentation.
Co-authored-by: @A-Hall
This removes the option to run a cluster without enforcing the
cluster-wide shard limit, making strict enforcement the default and only
behavior. The limit can still be adjusted as desired using the cluster
settings API.
Add GREATEST(expr1, expr2, ... exprN) and LEAST(expr1, expr2, exprN)
functions which are in the family of CONDITIONAL functions.
Implementation follows PostgreSQL behaviour, so the functions return
`NULL` when all of their arguments evaluate to `NULL`.
Renamed `CoalescePipe` and `CoalesceProcessor` to `ConditionalPipe` and
`ConditionalProcessor` respectively, to be able to reuse them for
`Greatest` and `Least` evaluations. To achieve that `ConditionalOperation`
has been added to differentiate between the functionalities at execution
time.
Closes: #35878
This operator handles nulls in different way than the normal `=`.
If one of the operants is `null` and the other not it returns `false`.
If both operants are `null` it returns `true`. Therefore in contrary to
`=`, which returns `null` if at least one of the operants is `null`, this one
never returns `null` as a result.
Closes: #35871
This endpoint was not previously documented as it was not
particularly useful to end users. However, since the HLRC
will support the endpoint we need some documentation to
link to.
The purpose of the endpoint is to provide defaults and
limits used by ML. These are needed to fully understand
configurations that have missing values because the missing
value means the default should be used.
Relates #35777
* Forbid negative scores in functon_score query
- Throw an exception when scores are negative in field_value_factor
function
- Throw an exception when scores are negative in script_score
function
Relates to #33309
This commit moves the documentation and examples for the `index_prefixes`
option on text fields to its own file, to bring it in line with other mapping
parameters, and expands a bit on both.
Recent Docker for Mac releases[1] have a different path to the tty for
accessing the console of the xhyve vm, required for altering the
`vm.max_map_count` sysctl.
Update instructions on how to enter the xhyve vm for altering the
`vm.max_map_count` sysctl setting on Docker for Mac.
Closes#34817
[1]
https://forums.docker.com/t/is-it-possible-to-ssh-to-the-xhyve-machine/17426/13
The ICU plugin provides the building blocks of an analysis chain, but doesn't actually have a prebuilt analyzer. It would be a better for users if there was a simple analyzer that they could use out of the box, and also something we can point to from the CJK Analyzer docs as a superior alternative.
Relates to #34285
This commit adds a rest endpoint for freezing and unfreezing an index.
Among other cleanups mainly fixing an issue accessing package private APIs
from a plugin that got caught by integration tests this change also adds
documentation for frozen indices.
Note: frozen indices are marked as `beta` and available as a basic feature.
Relates to #34352
This adds an example to drive home the semantics of `min_age` and explain
how actions from one phase must complete before `min_age` is even tested
closes#34020.
RolloverStep previously had a name of "attempt_rollover", which was
inconsistent with all other step names due it its use of an underscore
instead of a dash.
* Deprecate types in count requests.
* Move RestCountAction to the 'search' package.
* Deprecate types in multi search requests.
* Add tests for types deprecation in the _search endpoint.
RolloverAction will now periodically check the rollover conditions using
the Rollover API with the dry_run option as an AsyncWaitStep, then run
the rollover itself by calling the Rollover API with no conditions,
which will always roll over, as an AsyncActionStep. This will resolve
race condition issues in policies using RolloverAction.
* ML: Adding missing datacheck to datafeedjob
* Adding client side and docs
* Making adjustments to validations
* Making values default to on, having more sensible limits
* Intermittent commit, still need to figure out interval
* Adjusting delayed data check interval
* updating docs
* Making parameter Boolean, so it is nullable
* bumping bwc to 7 before backport
* changing to version current
* moving delayed data check config its own object
* Separation of duties for delayed data detection
* fixing checkstyles
* fixing checkstyles
* Adjusting default behavior so that null windows are allowed
* Mentioning the default value
* Fixing comments, syncing up validations
The docs are not resilient to timing issues where the ILM metadata is not set on newly
created indices, so we shouldn't be so strict on the returned response
The documentation of `search_after` recommends to use the `_id`
field as a tiebreaker for the sort without warning against
the additional memory required. This change changes the recommandation
to use a copy of the `_id` field with doc_values enabled.
This adds a `wait_for_completion` flag which allows the user to block
the Stop API until the task has actually moved to a stopped state,
instead of returning immediately. If the flag is set, a `timeout` parameter
can be specified to determine how long (at max) to block the API
call. If unspecified, the timeout is 30s.
If the timeout is exceeded before the job moves to STOPPED, a
timeout exception is thrown. Note: this is just signifying that the API
call itself timed out. The job will remain in STOPPING and evenutally
flip over to STOPPED in the background.
If the user asks the API to block, we move over the the generic
threadpool so that we don't hold up a networking thread.
* [ILM] Add documentation for error handling in ILM
This adds some initial documentation for error handling and retrying failed
steps for index lifecycle management
We changed the way realm settings are defined, and this affects custom
realms in SecurityExtensions. This change adds those details to the
breaking changes docs.
Relates: #30241
* [DOCS] ILM API Ref edits
* [DOCS] Fixed endpoint for DELETE policy.
* [DOCS] Removed comparison to setting index.lifecycle.name to null.
* [DOCS] Fixed xrefs to explain API.
Today our OS information returned in node stats only returns a
high-level name of the OS (e.g., "Linux"). Yet, for some uses this is
too high-level and knowing at a finer level of granularity the
underlying OS can be useful. This commit extracts the pretty name on
Linux from /etc/os-release. This pretty name usually includes the Linux
vendor and the Linux vendor version number (e.g., Fedora 28).
Currently we introduced a hard limit of 1024 to the number of fields a query can
be expanded to in #26541. Instead of using a hard limit, we should make this
configurable. This change removes the hard limit check and uses the existing
`max_clause_count` setting instead.
Closes#34778
If the underlying mount point for the JNA temporary directory is mounted
noexec on Linux, then the JVM will not be able to map the native code in
as executable. This will prevent JNA from executing and will prevent
Elasticsearch from being able to execute some functions that rely on
native code (e.g., memory locking, and installing system call
filters). We do not want to get into the business of catching exceptions
and parsing messages towards this because these exception messages can
change on us. We also do not want to jump through a lot of hoops to
check the underlying mount point for noexec. Instead, we will rely on
documentation to address this problem. This commit adds to the important
system configuration section of the docs that the JNA temporary
directory is not on a mount point with the noexec mount option.
This commit uses the index settings version so that a follower can
replicate index settings changes as needed from the leader.
Co-authored-by: Martijn van Groningen <martijn.v.groningen@gmail.com>
Sometimes users are confused about whether they can use the Convert Processor
for changing an existing fields type to other types even if the existing one is already
ingested. This confusion is from the first line of description. Changing this and also
adding a some detail to the code snippet.
With this change, `Version` no longer carries information about the qualifier,
we still need a way to show the "display version" that does have both
qualifier and snapshot. This is now stored by the build and red from `META-INF`.
We've decided that the bulk, delete, get, index, update, and search APIs should not
contain this request parameter, and we will instead accept both typed and typeless calls.
The remove-ilm-from-index API was using the DELETE http method
to signify that something is being removed. Although, metadata
about ILM for the index is being deleted, no entity/resource
is being deleted during this operation. POST is more in line with
what this API is actually doing, it is modifying the metadata for
an index. As part of this change, `remove` is also appended to the path
to be more explicit about its actions.
This moves all Realm settings to an Affix definition.
However, because different realm types define different settings
(potentially conflicting settings) this requires that the realm type
become part of the setting key.
Thus, we now need to define realm settings as:
xpack.security.authc.realms:
file.file1:
order: 0
native.native1:
order: 1
- This is a breaking change to realm config
- This is also a breaking change to custom security realms (SecurityExtension)
We have an example in `reindex`'s docs about copying from many indices
at once. It doesn't work at the moment because we only allow a single
type per index. We didn't notice it in the docs tests because those
tests didn't copy any documents. This change:
1. Adds documents to the docs tests to fully exercise the snippet.
2. Fixes the example by moving all copied documents to the same type.
3. Moves the note about id collisions and expands on it because it is
even more likely than before.
Closes#35150
This commit removes the Joda time usage from ILM and the HLRC components of ILM.
It also fixes an issue where using the `?human=true` flag could have caused the
parser not to work. These millisecond fields now follow the standard we use
elsewhere in the code, with additional fields added iff the `human` flag is
specified.
This is a breaking change for ILM, but since ILM has not yet been released, no
compatibility shim is needed.
This changes the current script.max_size_in_bytes to be dynamic so it can be
set through the cluster settings API. This setting is also applied to inline scripts
in the compile method of ScriptService to prevent excessively long inline
scripts from being compiled. The script length limit is removed from Painless as
this is no longer necessary with the protection in compile.
With this commit we differentiate between permanent circuit breaking
exceptions (which require intervention from an operator and should not
be automatically retried) and transient ones (which may heal themselves
eventually and should be retried). Furthermore, the parent circuit
breaker will categorize a circuit breaking exception as either transient
or permanent based on the categorization of memory usage of its child
circuit breakers.
Closes#31986
Relates #34460
When we connect to remote clusters, there may be a few more routers/firewalls in-between compared to when we connect to nodes in the same cluster. We've experienced cases where firewalls drop connections completely and keep-alives seem not to be enough, or they are not properly configured. With this commit we allow to enable application-level pings specifically from CCS nodes to the selected remote nodes through the new setting `cluster.remote.${clusterAlias}.transport.ping_schedule`. The new setting is similar `transport.ping_schedule` but it does not affect intra-cluster communication, pings are only sent to specific remote cluster when specifically enabled, as they are disabled by default.
Relates to #34405
This PR renames the CRUD APIS for ILM
GET _ilm/<policy>, _ilm -> _ilm/policy/<policy>, _ilm/policy
PUT _ilm/<policy> -> _ilm/policy/<policy>
DELETE _ilm/<policy> -> _ilm/policy/<policy>
closes#34929.
The `random_score` function produces values between 0 (inclusive) and 1
(exclusive) and documented it with fancy methematical range notation. It
is so fancy I thought it was a typo. This changes the documentation to
use words.
Relates to #35084
This changes the RollupSearch endpoint to proactively resolve index
patterns. If the index pattern(s) match more than one rollup index,
an exception is throw as before. But if the pattern only matches one
rollup index, execution is allowed to continue (unlike before where
it would assume all patterns were for raw data).
This also allows the search endpoint to resolve aliases that point to
a rollup index.
Also tweaks the documentation to make this clear.
Closes#34828
* Remove a tip about ignore_above that only makes sense with multiple types.
* Remove a line from the percolator documentation that refers to multiple types.
This commit adds a new single value metric aggregation that calculates
the statistic called median absolute deviation, which is a measure of
variability that works on more types of data than standard deviation
Our calculation of MAD is approximated using t-digests. In the collect
phase, we collect each value visited into a t-digest. In the reduce
phase, we merge all value t-digests, then create a t-digest of
deviations using the first t-digest's median and centroids
When combine_script and reduce_script were made into required
parameters for Scripted Metric aggregations in #33452, the docs were
not updated to reflect that. This marks those parameters as required
in the documentation.
Deprecates `_source_include` and `_source_exclude` url parameters
in favor of `_source_inclues` and `_source_excludes` because those
are consistent with the rest of Elasticsearch's APIs.
Relates to #22792
This commit fixes two issues with the CCR API specification:
- remove the CCR stats endpoint, it is not currently implemented
- fix the documentation links
The file structure finder endpoint can find the NDJSON
(newline-delimited JSON) file format, but called it
`json`. This change renames the `format` for this file
structure to `ndjson`, which is more precise and will
hopefully avoid confusion.
* Changed the auto follow stats to also include follow stats.
* Renamed the auto follow stats api to stats api and changed its url path
from `/_ccr/auto_follow/stats` `/_ccr/stats`.
* Removed `/_ccr/stats` url path for the follow stats api, which makes
the index parameter a required parameter.
* Fixed docs.
This commit is our first introduction to cross-cluster replication
docs. In this commit, we introduce the cross-cluster replication API
docs. We also add skelton docs for additional content that will be added
in a series of follow-up commits.
Documents the new structured logfile format for auditing
that was introduced by #31931. Most changes herein
are for 6.x . In 7.0 the deprecated format is gone and a
follow-up PR is in order.
This change adds a section about the global search setting
`indices.query.bool.max_clause_count` that limits the number of boolean clauses
allowed in a Lucene BooleanQuery.
Closes#19858
In a future major version, we will be introducing a soft limit on the
number of shards in a cluster based on the number of nodes in the
cluster. This limit will be configurable, and checked on operations
which create or open shards and issue a warning if the operation would
take the cluster over the limit.
There is an option to enable strict enforcement of the limit, which
turns the warnings into errors. In a future release, the option will be
removed and strict enforcement will be the default (and only) behavior.
- Restrict visibility of Aggregators and Factories
- Move PipelineAggregatorBuilders up a level so it is consistent with
AggregatorBuilders
- Checkstyle line length fixes for a few classes
- Minor odds/ends (swapping to method references, formatting, etc)
We should delete a job by directly talking to the allocated
task and telling it to shutdown. Today we shut down a job
via the persistent task framework. This is not ideal because,
while the job has been removed from the persistent task
CS, the allocated task continues to live until it gets the
shutdown message.
This means a user can delete a job, immediately delete
the rollup index, and then see new documents appear in
the just-deleted index. This happens because the indexer
in the allocated task is still running and indexes a few
more documents before getting the shutdown command.
In this PR, the transport action is changed to a TransportTasksAction,
and we invoke onCancelled() directly on the matching job.
The race condition still exists after this PR (albeit less likely),
but this was a precursor to fixing the issue and a self-contained
chunk of code. A second PR will followup to fix the race itself.
Extend querying support on multiple indices from being strictly
identical to being just compatible.
Use FieldCapabilities API (extended through #33803) for mapping merging.
Close#31837#31611
Implement the functionality to translate the
`field IN (value1, value2,...)` expressions to proper Lucene queries
or painless script or local processors depending on the use case.
The `IN` expression can be used in SELECT, WHERE and HAVING clauses.
Closes: #32955
`CONVERT` works exactly like cast with slightly different syntax:
`CONVERT(<value>, <data_type)` as opposed to `CAST(<value> AS <data_type>)`
Moreover it support format of the MS-SQL data types `SQL_<type>`,
e.g.: `SQL_INTEGER`
Closes: #34513
* Replace custom type names with _doc in REST examples.
* Avoid using two mapping types in the percolator docs.
* Rename doc -> _doc in the main repository README.
* Also replace some custom type names in the HLRC docs.
This commit switches to using a trial license in the docs tests that run
on the default distribution. This is needed so that docs tests can be
executed against non-basic features.
With remote clusters taking on a larger role, we have make the
infrastructure more generic than being tied to cross-cluster search
(CCS). We want to refer to the remote clusters configuration in the
cross-cluster replication (CCR) docs. Yet, these docs are still tied to
CCS. This commit extracts the remote clusters docs from CCS (with some
wording changes to make them more general) so that we can refer to them
in the CCR docs.
When a envelope that crosses the dateline is specified as a part of
geo_shape query is parsed it shouldn't have its left and right points
flipped.
Fixes#34418
Make SQL aware of missing and/or unmapped fields treating them as NULL
Make _all_ functions and operators null-safe aware, including when used
in filtering or sorting contexts
Add missing and null-safe doc value extractor
Modify dataset to have null fields spread around (in groups of 10)
Enforce missing last and unmapped_type inside sorting
Consolidate Predicate templating and declaration
Add support for Like/RLike in scripting
Generalize NULLS LAST/FIRST
Introduce early schema declaration for CSV spec tests: to keep the doc
snippets in place (introduce schema:: prefix for declaration)
upfront.
Fix#32079
The `term` and `phrase` suggesters have different options to filter candidates
based on their frequencies. The `popular` mode for instance filters candidate
terms that occur in less docs than the original term. However when we compute this threshold
we use the total term frequency of a term instead of the document frequency. This is not inline
with the actual filtering which is always based on the document frequency. This change fixes
this discrepancy and clarifies the meaning of the different frequencies in use in the suggesters.
It also ensures that the threshold doesn't overflow the maximum allowed value (Integer.MAX_VALUE).
Closes#34282
Add example for selectively clearing just the request, query or fielddata cache
and for selectively clearing the cache for specific fields.
Closes#34287
* Adding new xpack.ml.max_lazy_ml_nodes setting to docs
* Fixing docs, making it clearer what the setting does
* Adding note about external process need
We'd disabled them because we didn't have a way to clean up after each
test. I implemented #34342 which adds the clean ups so now we can
re-enable the tests.
In the `setup` sections we have to use `raw` requests instead of
`x-pack` requests because we don't have the json config for x-pack.
Closes#33319
This commit moves the definition of domainSplit into java and exposes it
as a painless whitelist extension. The method also no longer needs
params, and version which ignores params is added and deprecated.
Tweak the upgrade instructions for moving from pre-6.3-with-x-pack to
post-6.3-default distribution. Specifically, you have to remove the
x-pack plugin before upgrading because 6.4 doesn't understand how to
remove it.
Relates to #34307
This change disallows negative query boosts. Negative scores are not allowed in Lucene 8 so
it is easier to just disallow negative boosts entirely. We should also deprecate negative boosts
in 6x in order to ensure that users are aware when they'll upgrade to ES 7.
Relates #33309
The ingest pipeline that is produced is very simple. It
contains a grok processor if the format is semi-structured
text, a date processor if the format contains a timestamp,
and a remove processor if required to remove the interim
timestamp field parsed out of semi-structured text.
Eventually the UI should offer the option to customize the
pipeline with additional processors to perform other data
preparation steps before ingesting data to an index.
* New OCTET_LENGTH function
* Changed the way the FunctionRegistry stores functions, considering the alphabetic ordering by name
* Added documentation for the RANDOM function
Marios Trivyzas
Yu
Yannick Welsch
David Turner
Julie Tibshirani
Gordon Brown
Christoph Büscher
Simon Willnauer
David Roberts
Nik Everett
Armin Braun
Jason Tedor
Tal Levy
Lisa Cawley
Jim Ferenczi
Alan Woodward
Martijn van Groningen
Guido Lena Cota
João Barbosa
Chris Koehnke
Christophe Bismuth
patrykk21
Peter Dyson
Zachary Tong
Toby McLaughlin
Ryan Ernst
Alpar Torok
Andrei Stefan
Costin Leau
Zachary Tong
Jeff Hajewski
Yogesh Gaikwad
Jake Landis
Ed Savage
William Desportes
Mayya Sharipova
Dimitrios Liappis
Marios Trivyzas
Takuro Wada
Itamar Syn-Hershko
Benjamin Trent
debadair
Jun Ohtani
Thiago Souza
Lee Hinman
Albert Zaharovits
Tim Vernum
Jim Ferenczi
Dominik Stadler
Russ Cam
Chris Cho
Gytis Šk
Jeff Soloshy
Tim Vernum
Alexander Zhukov
Jack Conradson
Gordon Brown
Colin Goodheart-Smithe
Daniel Mitterdorfer
Alexandru Rusanescu
lipsill
Qian Hong
Luca Cavanna
Sue Gallagher
Andy Bristol
markwalkom
Ken MacInnis
Senthil
Stéphane Campinas
Lee Hinman
Przemyslaw Gomulka
Frederik Bosch
Igor Motov
markharwood
Nikolay Vasiliev
amoreauCoveo
eray
Ioannis Kakavas