When a local model is constructed, the cache hit miss count is incremented.
When a user calls _stats, we will include the sum cache hit miss count across ALL nodes. This statistic is important to in comparing against the inference_count. If the cache hit miss count is near the inference_count it indicates that the cache is overburdened, or inappropriately configured.
Deleting expired data can take a long time leading to timeouts if there
are many jobs. Often the problem is due to a few large jobs which
prevent the regular maintenance of the remaining jobs. This change adds
a job_id parameter to the delete expired data endpoint to help clean up
those problematic jobs.
This PR adds the initial Java side changes to enable
use of the per-partition categorization functionality
added in elastic/ml-cpp#1293.
There will be a followup change to complete the work,
as there cannot be any end-to-end integration tests
until elastic/ml-cpp#1293 is merged, and also
elastic/ml-cpp#1293 does not implement some of the
more peripheral functionality, like stop_on_warn and
per-partition stats documents.
The changes so far cover REST APIs, results object
formats, HLRC and docs.
Backport of #57683
When we force delete a DF analytics job, we currently first force
stop it and then we proceed with deleting the job config.
This may result in logging errors if the job config is deleted
before it is retrieved while the job is starting.
Instead of force stopping the job, it would make more sense to
try to stop the job gracefully first. So we now try that out first.
If normal stop fails, then we resort to force stopping the job to
ensure we can go through with the delete.
In addition, this commit introduces `timeout` for the delete action
and makes use of it in the child requests.
Backport of #57680
* [ML] adds new for_export flag to GET _ml/inference API (#57351)
Adds a new boolean flag, `for_export` to the `GET _ml/inference/<model_id>` API.
This flag is useful for moving models between clusters.
This adds a max_model_memory setting to forecast requests.
This setting can take a string value that is formatted according to byte sizes (i.e. "50mb", "150mb").
The default value is `20mb`.
There is a HARD limit at `500mb` which will throw an error if used.
If the limit is larger than 40% the anomaly job's configured model limit, the forecast limit is reduced to be strictly lower than that value. This reduction is logged and audited.
related native change: https://github.com/elastic/ml-cpp/pull/1238
closes: https://github.com/elastic/elasticsearch/issues/56420
Throttling nightly cleanup as much as we do has been over cautious.
Night cleanup should be more lenient in its throttling. We still
keep the same batch size, but now the requests per second scale
with the number of data nodes. If we have more than 5 data nodes,
we don't throttle at all.
Additionally, the API now has `requests_per_second` and `timeout` set.
So users calling the API directly can set the throttling.
This commit also adds a new setting `xpack.ml.nightly_maintenance_requests_per_second`.
This will allow users to adjust throttling of the nightly maintenance.
This PR implements the following changes to make ML model snapshot
retention more flexible in advance of adding a UI for the feature in
an upcoming release.
- The default for `model_snapshot_retention_days` for new jobs is now
10 instead of 1
- There is a new job setting, `daily_model_snapshot_retention_after_days`,
that defaults to 1 for new jobs and `model_snapshot_retention_days`
for pre-7.8 jobs
- For days that are older than `model_snapshot_retention_days`, all
model snapshots are deleted as before
- For days that are in between `daily_model_snapshot_retention_after_days`
and `model_snapshot_retention_days` all but the first model snapshot
for that day are deleted
- The `retain` setting of model snapshots is still respected to allow
selected model snapshots to be retained indefinitely
Backport of #56125
The failed_category_count statistic records the number of times
categorization wanted to create a new category but couldn't
because the job had reached its model_memory_limit.
Backport of #55716
The ML info endpoint returns the max_model_memory_limit setting
if one is configured. However, it is still possible to create
a job that cannot run anywhere in the current cluster because
no node in the cluster has enough memory to accommodate it.
This change adds an extra piece of information,
limits.effective_max_model_memory_limit, to the ML info
response that returns the biggest model memory limit that could
be run in the current cluster assuming no other jobs were
running.
The idea is that the ML UI will be able to warn users who try to
create jobs with higher model memory limits that their jobs will
not be able to start unless they add a bigger ML node to their
cluster.
Backport of #55529
Adds a "node" field to the response from the following endpoints:
1. Open anomaly detection job
2. Start datafeed
3. Start data frame analytics job
If the job or datafeed is assigned to a node immediately then
this field will return the ID of that node.
In the case where a job or datafeed is opened or started lazily
the node field will contain an empty string. Clients that want
to test whether a job or datafeed was opened or started lazily
can therefore check for this.
Backport of #55473
This paves the data layer way so that exceptionally large models are partitioned across multiple documents.
This change means that nodes before 7.8.0 will not be able to use trained inference models created on nodes on or after 7.8.0.
I chose the definition document limit to be 100. This *SHOULD* be plenty for any large model. One of the largest models that I have created so far had the following stats:
~314MB of inflated JSON, ~66MB when compressed, ~177MB of heap.
With the chunking sizes of `16 * 1024 * 1024` its compressed string could be partitioned to 5 documents.
Supporting models 20 times this size (compressed) seems adequate for now.
* [ML] adding prediction_field_type to inference config (#55128)
Data frame analytics dynamically determines the classification field type. This field type then dictates the encoded JSON that is written to Elasticsearch.
Inference needs to know about this field type so that it may provide the EXACT SAME predicted values as analytics.
Here is added a new field `prediction_field_type` which indicates the desired type. Options are: `string` (DEFAULT), `number`, `boolean` (where close_to(1.0) == true, false otherwise).
Analytics provides the default `prediction_field_type` when the model is created from the process.
* [ML] add new inference_config field to trained model config (#54421)
A new field called `inference_config` is now added to the trained model config object. This new field allows for default inference settings from analytics or some external model builder.
The inference processor can still override whatever is set as the default in the trained model config.
* fixing for backport
* [ML] prefer secondary authorization header for data[feed|frame] authz (#54121)
Secondary authorization headers are to be used to facilitate Kibana spaces support + ML jobs/datafeeds.
Now on PUT/Update/Preview datafeed, and PUT data frame analytics the secondary authorization is preferred over the primary (if provided).
closes https://github.com/elastic/elasticsearch/issues/53801
* fixing for backport
* [ML] add num_matches and preferred_to_categories to category defintion objects (#54214)
This adds two new fields to category definitions.
- `num_matches` indicating how many documents have been seen by this category
- `preferred_to_categories` indicating which other categories this particular category supersedes when messages are categorized.
These fields are only guaranteed to be up to date after a `_flush` or `_close`
native change: https://github.com/elastic/ml-cpp/pull/1062
* adjusting for backport
This is a simple naming change PR, to fix the fact that "metadata" is a
single English word, and for too long we have not followed general
naming conventions for it. We are also not consistent about it, for
example, METADATA instead of META_DATA if we were trying to be
consistent with MetaData (although METADATA is correct when considered
in the context of "metadata"). This was a simple find and replace across
the code base, only taking a few minutes to fix this naming issue
forever.
István Zoltán Szabó
Benjamin Trent
Przemysław Witek
David Kyle
David Roberts
Dimitris Athanasiou
Lisa Cawley
Jason Tedor