Scheduler #

Scheduler defines how often to run and make inferences, as well as what timerange to use to train the model. Is specified in scheduler section of a config for VictoriaMetrics Anomaly Detection.

Note: Scheduler section in config supports multiple schedulers via aliasing Available from v1.11.0 .
Also, vmanomaly expects scheduler section to be named schedulers. Using old (flat) format with scheduler key is deprecated and will be removed in future versions.

schedulers:
  scheduler_periodic_1m:
    # class: "periodic" # or class: "scheduler.periodic.PeriodicScheduler" until v1.13.0 with class alias support)
    infer_every: "1m"
    fit_every: "2m"
    fit_window: "3h"
  scheduler_periodic_5m:
    # class: "periodic" # or class: "scheduler.periodic.PeriodicScheduler" until v1.13.0 with class alias support)
    infer_every: "5m"
    fit_every: "10m"
    fit_window: "3h"
...

Old-style configs Deprecated from v1.11.0

scheduler:
  # class: "periodic" # or class: "scheduler.periodic.PeriodicScheduler" until v1.13.0 with class alias support)
  infer_every: "1m"
  fit_every: "2m"
  fit_window: "3h"
...

will be implicitly converted to

schedulers:
  default_scheduler:  # default scheduler alias added, for backward compatibility
    class: "scheduler.periodic.PeriodicScheduler"
    infer_every: "1m"
    fit_every: "2m"
    fit_window: "3h"
...

Parameters #

class: str, default="scheduler.periodic.PeriodicScheduler", options={"scheduler.periodic.PeriodicScheduler", "scheduler.oneoff.OneoffScheduler", "scheduler.backtesting.BacktestingScheduler"}

• "scheduler.periodic.PeriodicScheduler": periodically runs the models on new data. Useful for consecutive re-trainings to counter data drift and model degradation over time.
• "scheduler.oneoff.OneoffScheduler": runs the process once and exits. Useful for testing.
• "scheduler.backtesting.BacktestingScheduler": imitates consecutive backtesting runs of OneoffScheduler. Runs the process once and exits. Use to get more granular control over testing on historical data.

Note: Class aliases are supported Available from v1.13.0 , so "scheduler.periodic.PeriodicScheduler" can be substituted to "periodic", "scheduler.oneoff.OneoffScheduler" - to "oneoff", "scheduler.backtesting.BacktestingScheduler" - to "backtesting"

Depending on selected class, different parameters should be used

Periodic scheduler #

Parameters #

For periodic scheduler parameters are defined as differences in times, expressed in difference units, e.g. days, hours, minutes, seconds.

Examples: "50s", "4m", "3h", "2d", "1w".

        </td>
        <td>str, Optional</td>
        <td>

        </td>
        <td>str, Optional</td>
        <td>

Periodic scheduler config example #

schedulers:
  periodic_scheduler_alias:
    class: "periodic"
    # (or class: "scheduler.periodic.PeriodicScheduler" for versions before v1.13.0, without class alias support)
    fit_window: "14d" 
    infer_every: "1m" 
    fit_every: "1h"
    start_from: "20:00"  # If launched before 20:00 (local Kyiv time), the first run starts today at 20:00. Otherwise, it starts tomorrow at 20:00.
    tz: "Europe/Kyiv"  # Defaults to 'UTC' if not specified.

This configuration specifies that vmanomaly will calculate a 14-day time window from the time of fit_every call to train the model. Starting at 20:00 Kyiv local time today (or tomorrow if launched after 20:00), the model will be retrained every hour using the most recent 14-day window, which always includes an additional hour of new data. The time window remains strictly 14 days and does not extend with subsequent retrains. Additionally, vmanomaly will perform model inference every minute, processing newly added data points using the most recent model.

Oneoff scheduler #

Parameters #

For Oneoff scheduler timeframes can be defined in Unix time in seconds or ISO 8601 string format. ISO format supported time zone offset formats are:

• Z (UTC)
• ±HH:MM
• ±HHMM
• ±HH

If a time zone is omitted, a timezone-naive datetime is used.

Defining fitting timeframe #

Defining inference timeframe #

ISO format scheduler config example #

schedulers:
  oneoff_scheduler_alias:
    class: "oneoff"
    # (or class: "scheduler.oneoff.OneoffScheduler" until v1.13.0 with class alias support)
    fit_start_iso: "2022-04-01T00:00:00Z"
    fit_end_iso: "2022-04-10T00:00:00Z"
    infer_start_iso: "2022-04-11T00:00:00Z"
    infer_end_iso: "2022-04-14T00:00:00Z"

UNIX time format scheduler config example #

schedulers:
  oneoff_scheduler_alias:
    class: "oneoff"
    # (or class: "scheduler.oneoff.OneoffScheduler" until v1.13.0 with class alias support)
    fit_start_s: 1648771200
    fit_end_s: 1649548800
    infer_start_s: 1649635200
    infer_end_s: 1649894400

Backtesting scheduler #

Parameters #

As for Oneoff scheduler, timeframes can be defined in Unix time in seconds or ISO 8601 string format. ISO format supported time zone offset formats are:

• Z (UTC)
• ±HH:MM
• ±HHMM
• ±HH

If a time zone is omitted, a timezone-naive datetime is used.

Parallelization #

Defining overall timeframe #

This timeframe will be used for slicing on intervals (fit_window, infer_window == fit_every), starting from the latest available time point, which is to_* and going back, until no full fit_window + infer_window interval exists within the provided timeframe.

Defining training timeframe #

The same explicit logic as in Periodic scheduler

Defining inference timeframe #

In BacktestingScheduler, the inference window is implicitly defined as a period between 2 consecutive model fit_every runs. The latest inference window starts from to_s - fit_every and ends on the latest available time point, which is to_s. The previous periods for fit/infer are defined the same way, by shifting fit_every seconds backwards until we get the last full fit period of fit_window size, which start is >= from_s.

ISO format scheduler config example #

schedulers:
  backtesting_scheduler_alias:
    class: "backtesting"
    # (or class: "scheduler.backtesting.BacktestingScheduler" until v1.13.0 with class alias support)
    from_iso: '2021-01-01T00:00:00Z'
    to_iso: '2021-01-14T00:00:00Z'
    fit_window: 'P14D'
    fit_every: 'PT1H'
    n_jobs: 1  # default = 1 (sequential), set it up to # of CPUs for parallel execution

UNIX time format scheduler config example #

schedulers:
  backtesting_scheduler_alias:
    class: "backtesting"
    # (or class: "scheduler.backtesting.BacktestingScheduler" until v1.13.0 with class alias support)
    from_s: 167253120
    to_s: 167443200
    fit_window: '14d'
    fit_every: '1h'
    n_jobs: 1  # default = 1 (sequential), set it up to # of CPUs for parallel execution