Data frame analytics classification currently only supports 2 classes for the
dependent variable. We were checking that the field's cardinality is not higher
than 2 but we should also check it is not less than that as otherwise the process
fails.
Backport of #51232
The ID of the datafeed's associated job was being obtained
frequently by looking up the datafeed task in a map that
was being modified in other threads. This could lead to
NPEs if the datafeed stopped running at an unexpected time.
This change reduces the number of places where a datafeed's
associated job ID is looked up to avoid the possibility of
failures when the datafeed's task is removed from the map
of running tasks during multi-step operations in other
threads.
Fixes#51285
There are two edge cases that can be ran into when example input is matched in a weird way.
1. Recursion depth could continue many many times, resulting in a HUGE runtime cost. I put a limit of 10 recursions (could be adjusted I suppose).
2. If there are no "fixed regex bits", exploring the grok space would result in a fence-post error during runtime (with assertions turned off)
Allows ML datafeeds to work with time fields that have
the "date_nanos" type _and make use of the extra precision_.
(Previously datafeeds only worked with time fields that were
exact multiples of milliseconds. So datafeeds would work
with "date_nanos" only if the extra precision over "date" was
not used.)
Relates #49889
This change introduces a new feature for indices so that they can be
hidden from wildcard expansion. The feature is referred to as hidden
indices. An index can be marked hidden through the use of an index
setting, `index.hidden`, at creation time. One primary use case for
this feature is to have a construct that fits indices that are created
by the stack that contain data used for display to the user and/or
intended for querying by the user. The desire to keep them hidden is
to avoid confusing users when searching all of the data they have
indexed and getting results returned from indices created by the
system.
Hidden indices have the following properties:
* API calls for all indices (empty indices array, _all, or *) will not
return hidden indices by default.
* Wildcard expansion will not return hidden indices by default unless
the wildcard pattern begins with a `.`. This behavior is similar to
shell expansion of wildcards.
* REST API calls can enable the expansion of wildcards to hidden
indices with the `expand_wildcards` parameter. To expand wildcards
to hidden indices, use the value `hidden` in conjunction with `open`
and/or `closed`.
* Creation of a hidden index will ignore global index templates. A
global index template is one with a match-all pattern.
* Index templates can make an index hidden, with the exception of a
global index template.
* Accessing a hidden index directly requires no additional parameters.
Backport of #50452
If 1000 different category definitions are created for a job in
the first 100 buckets it processes then an audit warning will now
be created. (This will cause a yellow warning triangle in the
ML UI's jobs list.)
Such a large number of categories suggests that the field that
categorization is working on is not well suited to the ML
categorization functionality.
Object fields cannot be used as features. At the moment _explain
API includes them and even worse it allows it does not error when
an object field is excluded. This creates the expectation to the
user that all children fields will also be excluded while it's not
the case.
This commit omits object fields from the _explain API and also
adds an error if an object field is included or excluded.
Backport of #51115
There have been occasional failures, presumably due to
too many tests running in parallel, caused by jobs taking
around 15 seconds to open. (You can see the job open
successfully during the cleanup phase shortly after the
failure of the test in these cases.) This change increases
the wait time from 10 seconds to 20 seconds to reduce the
risk of this happening.
Check it out:
```
$ curl -u elastic:password -HContent-Type:application/json -XPOST localhost:9200/test/_update/foo?pretty -d'{
"dac": {}
}'
{
"error" : {
"root_cause" : [
{
"type" : "x_content_parse_exception",
"reason" : "[2:3] [UpdateRequest] unknown field [dac] did you mean [doc]?"
}
],
"type" : "x_content_parse_exception",
"reason" : "[2:3] [UpdateRequest] unknown field [dac] did you mean [doc]?"
},
"status" : 400
}
```
The tricky thing about implementing this is that x-content doesn't
depend on Lucene. So this works by creating an extension point for the
error message using SPI. Elasticsearch's server module provides the
"spell checking" implementation.
s
* [ML][Inference] Adding classification_weights to ensemble models
classification_weights are a way to allow models to
prefer specific classification results over others
this might be advantageous if classification value
probabilities are a known quantity and can improve
model error rates.
In classes where the client is used directly rather than through a call to
executeAsyncWithOrigin explicitly require the client to be OriginSettingClient
rather than using the Client interface.
Also remove calls to deprecated ClientHelper.clientWithOrigin() method.
Adds a new parameter to regression and classification that enables computation
of importance for the top most important features. The computation of the importance
is based on SHAP (SHapley Additive exPlanations) method.
Backport of #50914
The system created and models we provide now use the `_xpack` user for uniformity with our other features
The `PUT` action is now an admin cluster action
And XPackClient class now references the action instance.
* [ML][Inference] PUT API (#50852)
This adds the `PUT` API for creating trained models that support our format.
This includes
* HLRC change for the API
* API creation
* Validations of model format and call
* fixing backport
This commit removes validation logic of source and dest indices
for data frame analytics and replaces it with using the common
`SourceDestValidator` class which is already used by transforms.
This way the validations and their messages become consistent
while we reduce code.
This means that where these validations fail the error messages
will be slightly different for data frame analytics.
Backport of #50841
A very large number of recursive calls can cause a stack overflow
exception. This commit forks the recursive calls for non-async
processors. Once forked, each thread will handle at most 10
recursive calls to help keep the stack size and thread count
down to a reasonable size.
This adds the necessary named XContent classes to the HLRC for the lang ident model. This is so the HLRC can call `GET _ml/inference/lang_ident_model_1?include_definition=true` without XContent parsing errors.
The constructors are package private as since this classes are used exclusively within the pre-packaged model (and require the specific weights, etc. to be of any use).
This PR adds per-field metadata that can be set in the mappings and is later
returned by the field capabilities API. This metadata is completely opaque to
Elasticsearch but may be used by tools that index data in Elasticsearch to
communicate metadata about fields with tools that then search this data. A
typical example that has been requested in the past is the ability to attach
a unit to a numeric field.
In order to not bloat the cluster state, Elasticsearch requires that this
metadata be small:
- keys can't be longer than 20 chars,
- values can only be numbers or strings of no more than 50 chars - no inner
arrays or objects,
- the metadata can't have more than 5 keys in total.
Given that metadata is opaque to Elasticsearch, field capabilities don't try to
do anything smart when merging metadata about multiple indices, the union of
all field metadatas is returned.
Here is how the meta might look like in mappings:
```json
{
"properties": {
"latency": {
"type": "long",
"meta": {
"unit": "ms"
}
}
}
}
```
And then in the field capabilities response:
```json
{
"latency": {
"long": {
"searchable": true,
"aggreggatable": true,
"meta": {
"unit": [ "ms" ]
}
}
}
}
```
When there are no conflicts, values are arrays of size 1, but when there are
conflicts, Elasticsearch includes all unique values in this array, without
giving ways to know which index has which metadata value:
```json
{
"latency": {
"long": {
"searchable": true,
"aggreggatable": true,
"meta": {
"unit": [ "ms", "ns" ]
}
}
}
}
```
Closes#33267
Switch from a 32 bit Java hash to a 128 bit Murmur hash for
creating document IDs from by/over/partition field values.
The 32 bit Java hash was not sufficiently unique, and could
produce identical numbers for relatively common combinations
of by/partition field values such as L018/128 and L017/228.
Fixes#50613
The end offset of a tokenizer is supposed to point one past the
end of the input, not to the end character of the input. The
ml_classic tokenizer was erroneously doing the latter.
Sharing a random generator may cause test failures as non-threadsafe random generators are periodically utilized in tests (see: https://github.com/elastic/elasticsearch/issues/50651)
This change constructs a calls `Randomness.get()` within the `bulkIndexWithRetry` method so that the returned `Random` object is only used in a single thread. Before, the member variable could have been used between threads, which caused test failures.
* [ML][Inference] lang_ident model (#50292)
This PR contains a java port of Google's CLD3 compact NN model https://github.com/google/cld3
The ported model is formatted to fit within our inference model formatting and stored as a resource in the `:xpack:ml:` plugin and is under basic license.
The model is broken up into two major parts:
- Preprocessing through the custom embedding (based on CLD3's embedding layer)
- Pushing the embedded text through the two layers of fully connected shallow NN.
Main differences between this port and CLD3:
- We take advantage of Java's internal Unicode handling where possible (i.e. codepoints, characters, decoders, etc.)
- We do not trim down input text by removing duplicated tokens
- We do not encode doubles/floats as longs/integers.
Adds a `force` parameter to the delete data frame analytics
request. When `force` is `true`, the action force-stops the
jobs and then proceeds to the deletion. This can be used in
order to delete a non-stopped job with a single request.
Closes#48124
Backport of #50553
Eclipse 4.13 shows a type mismatch error in the affected line because it cannot
correctly infer the boolean return type for the method call. Assigning return
value to a local variable resolves this problem.
In order to ensure any persisted model state is searchable by the moment
the job reports itself as `stopped`, we need to refresh the state index
before completing.
This should fix the occasional failures we see in #50168 and #50313 where
the model state appears missing.
Closes#50168Closes#50313
Backport of #50322
This fixes support for nested fields
We now support fully nested, fully collapsed, or a mix of both on inference docs.
ES mappings allow the `_source` to be any combination of nested objects + dot delimited fields.
So, we should do our best to find the best path down the Map for the desired field.
This commit adds removal of unused data frame analytics state
from the _delete_expired_data API (and in extend th ML daily
maintenance task). At the moment the potential state docs
include the progress document and state for regression and
classification analyses.
Backport of #50243
Executing the data frame analytics _explain API with a config that contains
a field that is not in the includes list but at the same time is the excludes
list results to trying to remove the field twice from the iterator. That causes
an `IllegalStateException`. This commit fixes this issue and adds a test that
captures the scenario.
Backport of #50192
This adds a new field for the inference processor.
`warning_field` is a place for us to write warnings provided from the inference call. When there are warnings we are not going to write an inference result. The goal of this is to indicate that the data provided was too poor or too different for the model to make an accurate prediction.
The user could optionally include the `warning_field`. When it is not provided, it is assumed no warnings were desired to be written.
The first of these warnings is when ALL of the input fields are missing. If none of the trained fields are present, we don't bother inferencing against the model and instead provide a warning stating that the fields were missing.
Also, this adds checks to not allow duplicated fields during processor creation.
This exchanges the direct use of the `Client` for `ResultsPersisterService`. State doc persistence will now retry. Failures to persist state will still not throw, but will be audited and logged.
This commit fixes a bug that caused the data frame analytics
_explain API to time out in a multi-node setup when the source
index was missing. When we try to create the extracted fields detector,
we check the index settings. If the index is missing that responds
with a failure that could be wrapped as a remote exception.
While we unwrapped correctly to check if the cause was an
`IndexNotFoundException`, we then proceeded to cast the original
exception instead of the cause.
Backport of #50176
* [ML] Add graceful retry for anomaly detector result indexing failures (#49508)
All results indexing now retry the amount of times configured in `xpack.ml.persist_results_max_retries`. The retries are done in a semi-random, exponential backoff.
* fixing test
The `ClassificationIT.testTwoJobsWithSameRandomizeSeedUseSameTrainingSet`
test was previously set up to just have 10 rows. With `training_percent`
of 50%, only 5 rows will be used for training. There is a good chance that
all 5 rows will be of one class which results to failure.
This commit increases the rows to 100. Now 50 rows should be used for training
and the chance of failure should be very small.
Backport of #50072
Adjusts the subclasses of `TransportMasterNodeAction` to use their own loggers
instead of the one for the base class.
Relates #50056.
Partial backport of #46431 to 7.x.
This adds a new `randomize_seed` for regression and classification.
When not explicitly set, the seed is randomly generated. One can
reuse the seed in a similar job in order to ensure the same docs
are picked for training.
Backport of #49990
When checking the cardinality of a field, the query should be take into account. The user might know about some bad data in their index and want to filter down to the target_field values they care about.
Work in progress in the c++ side is increasing memory estimates
a bit and this test fails. At the time of this commit the mem
estimate when there is no source query is a about 2Mb. So I
am relaxing the test to assert memory estimate is less than 1Mb
instead of 500Kb.
Backport of #49924
This adds a `_source` setting under the `source` setting of a data
frame analytics config. The new `_source` is reusing the structure
of a `FetchSourceContext` like `analyzed_fields` does. Specifying
includes and excludes for source allows selecting which fields
will get reindexed and will be available in the destination index.
Closes#49531
Backport of #49690
We depend on the number of data frame rows in order to report progress
for the writing of results, the last phase of a job run. However, results
include other objects than just the data frame rows (e.g, progress, inference model, etc.).
The problem this commit fixes is that if we receive the last data frame row results
we'll report that progress is complete even though we still have more results to process
potentially. If the job gets stopped for any reason at this point, we will not be able
to restart the job properly as we'll think that the job was completed.
This commit addresses this by limiting the max progress we can report for the
writing_results phase before the results processor completes to 98.
At the end, when the process is done we set the progress to 100.
The commit also improves failure capturing and reporting in the results processor.
Backport of #49551
The categorization job wizard in the ML UI will use this
information when showing the effect of the chosen categorization
analyzer on a sample of input.
Before this change excluding an unsupported field resulted in
an error message that explained the excluded field could not be
detected as if it doesn't exist. This error message is confusing.
This commit commit changes this so that there is no error in this
scenario. When excluding a field that does exist but has been
automatically been excluded from the analysis there is no harm
(unlike excluding a missing field which could be a typo).
Backport of #49535
This commit replaces the _estimate_memory_usage API with
a new API, the _explain API.
The API consolidates information that is useful before
creating a data frame analytics job.
It includes:
- memory estimation
- field selection explanation
Memory estimation is moved here from what was previously
calculated in the _estimate_memory_usage API.
Field selection is a new feature that explains to the user
whether each available field was selected to be included or
not in the analysis. In the case it was not included, it also
explains the reason why.
Backport of #49455
If a datafeed is stopped normally and force stopped at the same
time then it is possible that the force stop removes the
persistent task while the normal stop is performing actions.
Currently this causes the normal stop to error, but since
stopping a stopped datafeed is not an error this doesn't make
sense. Instead the force stop should just take precedence.
This is a followup to #49191 and should really have been
included in the changes in that PR.
This commit moves the async calls required to retrieve the components
that make up `ExtractedFieldsExtractor` out of `DataFrameDataExtractorFactory`
and into a dedicated `ExtractorFieldsExtractorFactory` class.
A few more refactorings are performed:
- The detector no longer needs the results field. Instead, it knows
whether to use it or not based on whether the task is restarting.
- We pass more accurately whether the task is restarting or not.
- The validation of whether fields that have a cardinality limit
are valid is now performed in the detector after retrieving the
respective cardinalities.
Backport of #49315
The following edge cases were fixed:
1. A request to force-stop a stopping datafeed is no longer
ignored. Force-stop is an important recovery mechanism
if normal stop doesn't work for some reason, and needs
to operate on a datafeed in any state other than stopped.
2. If the node that a datafeed is running on is removed from
the cluster during a normal stop then the stop request is
retried (and will likely succeed on this retry by simply
cancelling the persistent task for the affected datafeed).
3. If there are multiple simultaneous force-stop requests for
the same datafeed we no longer fail the one that is
processed second. The previous behaviour was wrong as
stopping a stopped datafeed is not an error, so stopping
a datafeed twice simultaneously should not be either.
Backport of #49191
* [ML] ML Model Inference Ingest Processor (#49052)
* [ML][Inference] adds lazy model loader and inference (#47410)
This adds a couple of things:
- A model loader service that is accessible via transport calls. This service will load in models and cache them. They will stay loaded until a processor no longer references them
- A Model class and its first sub-class LocalModel. Used to cache model information and run inference.
- Transport action and handler for requests to infer against a local model
Related Feature PRs:
* [ML][Inference] Adjust inference configuration option API (#47812)
* [ML][Inference] adds logistic_regression output aggregator (#48075)
* [ML][Inference] Adding read/del trained models (#47882)
* [ML][Inference] Adding inference ingest processor (#47859)
* [ML][Inference] fixing classification inference for ensemble (#48463)
* [ML][Inference] Adding model memory estimations (#48323)
* [ML][Inference] adding more options to inference processor (#48545)
* [ML][Inference] handle string values better in feature extraction (#48584)
* [ML][Inference] Adding _stats endpoint for inference (#48492)
* [ML][Inference] add inference processors and trained models to usage (#47869)
* [ML][Inference] add new flag for optionally including model definition (#48718)
* [ML][Inference] adding license checks (#49056)
* [ML][Inference] Adding memory and compute estimates to inference (#48955)
* fixing version of indexed docs for model inference
This commit fixes a NPE problem as reported in #49150.
But this problem uncovered that we never added proper handling
of state for data frame analytics tasks.
In this commit we improve the `MlTasks.getDataFrameAnalyticsState`
method to handle null tasks and state tasks properly.
Closes#49150
Backport of #49186
Backport of #48849. Update `.editorconfig` to make the Java settings the
default for all files, and then apply a 2-space indent to all `*.gradle`
files. Then reformat all the files.
In the case multi-fields exist in the source index, we pick
all variants of them in our extracted fields detection for
data frame analytics. This means we may have multiple instances
of the same feature. The worse consequence of this is when the
dependent variable (for regression or classification) is also
duplicated which means we train a model on the dependent variable
itself.
Now that #48770 is merged, this commit is adding logic to
only select one variant of multi-fields.
Closes#48756
Backport of #48799
Aggregatable mutli-fields are at the moment wrongly mapped
as normal doc_value fields and thus they support fetching from
source. However, they do not exist in the source. This results
to failure to extract such fields.
This commit fixes this bug. While a fix could be worked out
on top of the existing code, it is evident the extraction logic
has become difficult to understand and maintain. As we also
want to deduplicate multi-fields for data frame analytics,
it seemed appropriate to refactor the code to simplify and
better handle the extraction of multi-fields.
Relates #48756
Backport of #48770
At present the ML C++ artifact is always downloaded from
S3. This change adds an option to configure the location.
(The intention is to use a file:/// URL to pick up the
artifact built in a Docker container in ml-cpp PR builds
so that C++ changes that will break Java integration tests
can be detected before the ml-cpp PRs are merged.)
Relates elastic/ml-cpp#766
Moves common field extraction logic to its own package so that it can
be used both for anomaly detection and data frame analytics.
In preparation for refactoring extraction fields to be simpler and to
support multi-fields properly.
Backport of #48709
* [ML][Inference] separating definition and config object storage (#48651)
This separates out the `definition` object from being stored within the configuration object in the index.
This allows us to gather the config object without decompressing a potentially large definition.
Additionally, `input` is moved to the TrainedModelConfig object and out of the definition. This is so the trained input fields are accessible outside the potentially large model definition.
There is a watchdog in order to avoid long running (and expensive)
grok expressions. Currently the watchdog is thread based, threads
that run grok expressions are registered and after completion unregister.
If these threads stay registered for too long then the watch dog interrupts
these threads. Joni (the library that powers grok expressions) has a
mechanism that checks whether the current thread is interrupted and
if so abort the pattern matching.
Newer versions have an additional method to abort long running pattern
matching inside joni. Instead of checking the thread's interrupted flag,
joni now also checks a volatile field that can be set via a `Matcher`
instance. This is more efficient method for aborting long running matches.
(joni checks each 30k iterations whether interrupted flag is set vs.
just checking a volatile field)
Recently we upgraded to a recent joni version (#47374), and this PR
is a followup of that PR.
This change should also fix#43673, since it appears when unit tests
are ran the a test runner thread's interrupted flag may already have
been set, due to some thread reuse.
Reverting the change introducing IsoLocal.ROOT and introducing IsoCalendarDataProvider that defaults start of the week to Monday and requires minimum 4 days in first week of a year. This extension is using java SPI mechanism and defaults for Locale.ROOT only.
It require jvm property java.locale.providers to be set with SPI,COMPAT
closes#41670
backport #48209
If a job stops right after reindexing is finished but before
we refresh the destination index, we don't refresh at all.
If the job is started again right after, it jumps into the analyzing state.
However, the data is still not searchable.
This is why we were seeing test failures that we start the process
expecting X rows (where X is lower than the expected number of docs)
and we end up getting X+.
We fix this by moving the refresh of the dest index right before
we start the process so it always ensures the data is searchable.
Closes#47612
Backport of #48090
This adds parsing an inference model as a possible
result of the analytics process. When we do parse such a model
we persist a `TrainedModelConfig` into the inference index
that contains additional metadata derived from the running job.
Audit messages are stored with millisecond timestamps. If two
messages have the same millisecond timestamp then asserting on
their order is impossible given the information available.
This PR changes the assertion on audit messages in the native
data frame analytics tests to assert that the expected audit
messages exist in any order.
Fixes#48035
This change adds:
- A new option, allow_lazy_open, to anomaly detection jobs
- A new option, allow_lazy_start, to data frame analytics jobs
Both work in the same way: they allow a job to be
opened/started even if no ML node exists that can
accommodate the job immediately. In this situation
the job waits in the opening/starting state until ML
node capacity is available. (The starting state for data
frame analytics jobs is new in this change.)
Additionally, the ML nightly maintenance tasks now
creates audit warnings for ML jobs that are unassigned.
This means that jobs that cannot be assigned to an ML
node for a very long time will show a yellow warning
triangle in the UI.
A final change is that it is now possible to close a job
that is not assigned to a node without using force.
This is because previously jobs that were open but
not assigned to a node were an aberration, whereas
after this change they'll be relatively common.
Adds a new datafeed config option, max_empty_searches,
that tells a datafeed that has never found any data to stop
itself and close its associated job after a certain number
of real-time searches have returned no data.
Backport of #47922
Usually syslog timestamps have two spaces before a single
digit day-of-month. However, in some non-syslog cases
where syslog-like timestamps are used there is only one
space. The grok pattern supports this, so the timestamp
parser should too. This change makes the
find_file_structure endpoint do this.
Also fixes another problem that the same test case
exposed in the find_file_structure endpoint, which was
that the exclude_lines_pattern for delimited files was
always created on the assumption the delimiter was a
comma. Now it is based on the actual delimiter.
* [ML][Analytics] fix bug where regression deleted early does not delete state (#47885)
* [ML][Analytics] fix bug where regression deleted early does not delete state
* Fixing ml with security test failure
* fixing for older java
When exceptions could be returned from another node, the exception
might be wrapped in a `RemoteTransportException`. In places where
we handled specific exceptions using `instanceof` we ought to unwrap
the cause first.
This commit attempts to fix this issue after searching code in the ML
plugin.
Backport of #47676
Adds the following parameters to `outlier_detection`:
- `compute_feature_influence` (boolean): whether to compute or not
feature influence scores
- `outlier_fraction` (double): the proportion of the data set assumed
to be outlying prior to running outlier detection
- `standardization_enabled` (boolean): whether to apply standardization
to the feature values
Backport of #47600
When an ML job runs the memory required can be
broken down into:
1. Memory required to load the executable code
2. Instrumented model memory
3. Other memory used by the job's main process or
ancilliary processes that is not instrumented
Previously we added a simple fixed overhead to
account for 1 and 3. This was 100MB for anomaly
detection jobs (large because of the completely
uninstrumented categorization function and
normalize process), and 20MB for data frame
analytics jobs.
However, this was an oversimplification because
the executable code only needs to be loaded once
per machine. Also the 100MB overhead for anomaly
detection jobs was probably too high in most cases
because categorization and normalization don't use
_that_ much memory.
This PR therefore changes the calculation of memory
requirements as follows:
1. A per-node overhead of 30MB for _only_ the first
job of any type to be run on a given node - this
is to account for loading the executable code
2. The established model memory (if applicable) or
model memory limit of the job
3. A per-job overhead of 10MB for anomaly detection
jobs and 5MB for data frame analytics jobs, to
account for the uninstrumented memory usage
This change will enable more jobs to be run on the
same node. It will be particularly beneficial when
there are a large number of small jobs. It will
have less of an effect when there are a small number
of large jobs.
While it seemed like the PUT data frame analytics action did not
have to be a master node action as the config is stored in an index
rather than the cluster state, there are other subtle nuances which
make it worthwhile to convert it. In particular, it helps maintain
order of execution for put actions which are anyhow user driven and
are expected to have low volume.
This commit converts `TransportPutDataFrameAnalyticsAction` from
a handled transport action to a master node action.
Note this means that the action might fail in a mixed cluster
but as the API is still experimental and not widely used there will
be few moments more suitable to make this change than now.
Due to #47003 many clusters will have built up a
large backlog of expired results. On upgrading to
a version where that bug is fixed users could find
that the first ML daily maintenance task deletes
a very large amount of documents.
This change introduces throttling to the
delete-by-query that the ML daily maintenance uses
to delete expired results to limit it to deleting an
average 200 documents per second. (There is no
throttling for state/forecast documents as these
are expected to be lower volume.)
Additionally a rough time limit of 8 hours is applied
to the whole delete expired data action. (This is only
rough as it won't stop part way through a single
operation - it only checks the timeout between
operations.)
Relates #47103
This commit restores the model state if available in data
frame analytics jobs.
In addition, this changes the start API so that a stopped job
can be restarted. As we now store the progress in the state index
when the task is stopped, we can use it to determine what state
the job was in when it got stopped.
Note that in order to be able to distinguish between a job
that runs for the first time and another that is restarting,
we ensure reindexing progress is reported to be at least 1
for a running task.
* [ML][Inference] adding .ml-inference* index and storage (#47267)
* [ML][Inference] adding .ml-inference* index and storage
* Addressing PR comments
* Allowing null definition, adding validation tests for model config
* fixing line length
* adjusting for backport
A refactoring in 6.6 meant that the ML daily
maintenance actions have not been run at all
since then. This change installs the local
master listener that schedules the ML daily
maintenance, and also defends against some
subtle race conditions that could occur in the
future if a node flipped very quickly between
master and non-master.
Fixes#47003
Backport of #45794 to 7.x. Convert most `awaitBusy` calls to
`assertBusy`, and use asserts where possible. Follows on from #28548 by
@liketic.
There were a small number of places where it didn't make sense to me to
call `assertBusy`, so I kept the existing calls but renamed the method to
`waitUntil`. This was partly to better reflect its usage, and partly so
that anyone trying to add a new call to awaitBusy wouldn't be able to find
it.
I also didn't change the usage in `TransportStopRollupAction` as the
comments state that the local awaitBusy method is a temporary
copy-and-paste.
Other changes:
* Rework `waitForDocs` to scale its timeout. Instead of calling
`assertBusy` in a loop, work out a reasonable overall timeout and await
just once.
* Some tests failed after switching to `assertBusy` and had to be fixed.
* Correct the expect templates in AbstractUpgradeTestCase. The ES
Security team confirmed that they don't use templates any more, so
remove this from the expected templates. Also rewrite how the setup
code checks for templates, in order to give more information.
* Remove an expected ML template from XPackRestTestConstants The ML team
advised that the ML tests shouldn't be waiting for any
`.ml-notifications*` templates, since such checks should happen in the
production code instead.
* Also rework the template checking code in `XPackRestTestHelper` to give
more helpful failure messages.
* Fix issue in `DataFrameSurvivesUpgradeIT` when upgrading from < 7.4
When the ML native multi-node tests use _cat/indices/_all
and the request goes to a non-master node, _all is
translated to a list of concrete indices by the authz layer
on the coordinating node before the request is forwarded
to the master node. Then it is possible for the master
node to return an index_not_found_exception if one of
the concrete indices that was expanded on the
coordinating node has been deleted in the meantime.
(#47159 has been opened to track the underlying problem.)
It has been observed that the index that gets deleted when
the problem affects the ML native multi-node tests is
always the ML notifications index. The tests that fail are
only interested in the presence or absense of ML results
indices. Therefore the workaround is to only _cat indices
that match the ML results index pattern.
Fixes#45652
* [ML][Inference] Feature pre-processing objects and functions (#46777)
To support inference on pre-trained machine learning models, some basic feature encoding will be necessary. I am using a named object serialization approach so new encodings/pre-processing steps could be added in the future.
This PR lays down the ground work for 3 basic encodings:
* HotOne
* Target Mean
* Frequency
More feature encodings or pre-processings could be added in the future:
* Handling missing columns
* Standardization
* Label encoding
* etc....
* fixing compilation for namedxcontent tests
When using auto-generated IDs + the ingest drop processor (which looks to be used by filebeat
as well) + coordinating nodes that do not have the ingest processor functionality, this can lead
to a NullPointerException.
The issue is that markCurrentItemAsDropped() is creating an UpdateResponse with no id when
the request contains auto-generated IDs. The response serialization is lenient for our
REST/XContent format (i.e. we will send "id" : null) but the internal transport format (used for
communication between nodes) assumes for this field to be non-null, which means that it can't
be serialized between nodes. Bulk requests with ingest functionality are processed on the
coordinating node if the node has the ingest capability, and only otherwise sent to a different
node. This means that, in order to reproduce this, one needs two nodes, with the coordinating
node not having the ingest functionality.
Closes#46678
This commit reuses the same state processor that is used for autodetect
to parse state output from data frame analytics jobs. We then index the
state document into the state index.
Backport of #46804
It is possible for a running analytics job that its config is removed
from the '.ml-config' index (perhaps the user deleted the entire index,
etc.). In that case the task remains without a matching config. I have
raised #46781 to discuss how to deal with this issue.
This commit focuses on `MlMemoryTracker` and changes it so that when
we get the configs for the running tasks we leniently ignore missing ones.
This at least means memory tracking will keep working for other jobs
if one or more are missing.
In addition, this commit makes the cleanup code for native analytics
tests more robust by explicitly stopping all jobs and force-stopping
if an error occurs. This helps so that a single failing test does
not cause other tests fail due to pending tasks.
Backport of #46789
When the stop API is called while the task is running there is
a chance the task gets marked completed twice. This may cause
undesired side effects, like indexing the progress document a second
time after the stop API has returned (the cause for #46705).
This commit adds a check that the task has not been completed before
proceeding to mark it so. In addition, when we update the task's state
we could get some warnings that the task was missing if the stop API
has been called in the meantime. We now check the errors are
`ResourceNotFoundException` and ignore them if so.
Closes#46705
Backports #46721
This is fixing a bug where if an analytics job is started before any
anomaly detection job is opened, we create an index after the state
write alias.
Instead, we should create the state index and alias before starting
an analytics job and this commit makes sure this is the case.
Backport of #46602
After starting the analytics job and checking its state
the state can be any of "started", "reindexing" or
"analyzing" depending on how quickly the work is done.
Investigating the test failure reported in #45518 it appears that
the datafeed task was not found during a tast state update. There
are only two places where such an update is performed: when we set
the state to `started` and when we set it to `stopping`. We handle
`ResourceNotFoundException` in the latter but not in the former.
Thus the test reveals a rare race condition where the datafeed gets
requested to stop before we managed to update its state to `started`.
I could not reproduce this scenario but it would be my best guess.
This commit catches `ResourceNotFoundException` while updating the
state to `started` and lets the task terminate smoothly.
Closes#45518
Backport of #46495
ML users who upgrade from versions prior to 7.4 to 7.4 or later
will have ML results indices that do not have mappings for the
total_search_time_ms field. Therefore, when searching these
indices we must tolerate this field not having a mapping.
Fixes#46437
This refactors `DataFrameAnalyticsTask` into its own class.
The task has quite a lot of functionality now and I believe it would
make code more readable to have it live as its own class rather than
an inner class of the start action class.
Backport of #46402
* [ML] waiting for ml indices before waiting task assignment testFullClusterRestart
* waiting for a stable cluster after fullrestart
* removing unused imports
The test seems to have been failing due to a race condition between
stopping the task and refreshing the destination index. In particular,
we were going forward with refreshing the destination index even
though the task stopped in the meantime. This was fixed in
request.
Closes#43960
Backport of #46271
Though we allow CCS within datafeeds, users could prevent nodes from accessing remote clusters. This can cause mysterious errors and difficult to troubleshoot.
This commit adds a check to verify that `cluster.remote.connect` is enabled on the current node when a datafeed is configured with a remote index pattern.
* [ML] Regression dependent variable must be numeric
This adds a validation that the dependent variable of a regression
analysis must be numeric.
* Address review comments and fix some problems
In addition to addressing the review comments, this
commit fixes a few issues I found during testing.
In particular:
- if there were mappings for required fields but they were
not included we were not reporting the error
- if explicitly included fields had unsupported types we were
not reporting the error
Unfortunately, I couldn't get those fixed without refactoring
the code in `ExtractedFieldsDetector`.
This commit adds support for `boolean` fields in data frame
analytics (and currently both outlier detection and regression).
The analytics process expects `boolean` fields to be encoded as
integers with 0 or 1 value.
Adds a parameter `training_percent` to regression. The default
value is `100`. When the parameter is set to a value less than `100`,
from the rows that can be used for training (ie. those that have a
value for the dependent variable) we randomly choose whether to actually
use for training. This enables splitting the data into a training set and
the rest, usually called testing, validation or holdout set, which allows
for validating the model on data that have not been used for training.
Technically, the analytics process considers as training the data that
have a value for the dependent variable. Thus, when we decide a training
row is not going to be used for training, we simply clear the row's
dependent variable.
The native process requires that there be a non-zero number of rows to analyze. If the flag --rows 0 is passed to the executable, it throws and does not start.
When building the configuration for the process we should not start the native process if there are no rows.
Adding some logging to indicate what is occurring.
Previously, the stats API reports a progress percentage
for DF analytics tasks that are running and are in the
`reindexing` or `analyzing` state.
This means that when the task is `stopped` there is no progress
reported. Thus, one cannot distinguish between a task that never
run to one that completed.
In addition, there are blind spots in the progress reporting.
In particular, we do not account for when data is loaded into the
process. We also do not account for when results are written.
This commit addresses the above issues. It changes progress
to being a list of objects, each one describing the phase
and its progress as a percentage. We currently have 4 phases:
reindexing, loading_data, analyzing, writing_results.
When the task stops, progress is persisted as a document in the
state index. The stats API now reports progress from in-memory
if the task is running, or returns the persisted document
(if there is one).
* [ML] Adding data frame analytics stats to _usage API (#45820)
* [ML] Adding data frame analytics stats to _usage API
* making the size of analytics stats 10k
* adjusting backport
Regression analysis support missing fields. Even more, it is expected
that the dependent variable has missing fields to the part of the
data frame that is not for training.
This commit allows to declare that an analysis supports missing values.
For such analysis, rows with missing values are not skipped. Instead,
they are written as normal with empty strings used for the missing values.
This also contains a fix to the integration test.
Closes#45425
* [ML] better handle empty results when evaluating regression
* adding new failure test to ml_security black list
* fixing equality check for regression results
We cannot know how long the analysis will take to complete thus we should not have
a timeout. Note that if the process crashes, the result processor will pick the
exception due to the stream closing.
Closes#45723
Changes the order of parameters in Geometries from lat, lon to lon, lat
and moves all Geometry classes are moved to the
org.elasticsearch.geomtery package.
Backport of #45332Closes#45048
* Reenable Integ Tests in native-multi-node-tests
* The tests broken here were likely fixed by #45463 => let's reenable them and see if things run fine again
* Relates #45405, #45455
This commit adds a first draft of a regression analysis
to data frame analytics. There is high probability that
the exact syntax might change.
This commit adds the new analysis type and its parameters as
well as appropriate validation. It also modifies the extractor
and the fields detector to be able to handle categorical fields
as regression analysis supports them.
In the FIPS JVM the JVM default locale seems to leak into places
where it should be overridden. This change skips assertions
in TimestampFormatFinderTests.testGuessIsDayFirstFromLocale
that may be impacted.
Fixes#45140
When doing a fieldwise Levenshtein distance comparison
between CSV rows, this change ignores all fields that
have long values, not just the longest field.
This approach works better for CSV formats that have
multiple freeform text fields rather than just a single
"message" field.
Fixes#45047
If one tries to start a DF analytics job that has already run,
the result will be that the task will fail after reindexing the
dest index from the source index. The results of the prior run
will be gone and the task state is not properly set to failed
with the failure reason.
This commit improves the behavior in this scenario. First, we
set the task state to `failed` in a set of failures that were
missed. Second, a validation is added that if the destination
index exists, it must be empty.
In case closing the process throws an exception we should be catching
it no matter its type. The process may have terminated because of a
fatal error in which case closing the process will throw a server
error, not an `IOException`. If this happens we fail to mark the
persistent task as failed and the task gets in limbo.
As data frame rows with missing values for analyzed fields are skipped,
we can be more efficient by including a query that only picks documents
that have values for all analyzed fields. Besides improving the number
of documents we go through, we also provide a more accurate measurement
of how many rows we need which reduces the memory requirements.
This also adds an integration test that runs outlier detection on data
with missing fields.
TaskListener accepts today Throwable in its onFailure method. Though
looking at where it is called (TransportAction), it can never be
notified of a Throwable.
This commit changes the signature of TaskListener#onFailure so that it
accepts an `Exception` rather than a `Throwable` as second argument.
David Kyle
Dimitris Athanasiou
Benjamin Trent
David Roberts
Przemysław Witek
Jay Modi
Tom Veasey
Nik Everett
Jake Landis
Adrien Grand
Christoph Büscher
David Turner
Rory Hunter
Martijn van Groningen
Przemyslaw Gomulka
rsarawgi
Igor Motov
Tanguy Leroux
Yannick Welsch
Armin Braun
Mark Vieira
Jason Tedor
Luca Cavanna