The error detail looks like:

  {
    'cause': 'Invalid input',
    'reason': 'boulet',
    'detail': ['field'],
  }

Note:

Voluptuous prints `r` before data path for
str(voluptuous.Invalid) in Python2. To simplify
tests in Python2 and Python3, the error reasons
are checked by regex.

This enables gnocchi-amqp daemon.

Which listens to metrics from collectd amqp1 write plugin.

support ability to groupby year, half, quarter, month, and week.

Some filesystem have a limit in term of number of directory a directory
can have. Also in performance point of view, some filesystem are slower
when too much directories are created in one directory.

To improve the situation is such case, this change adds an option
to create subdirectories for storing metrics.

for example file_subdir_len=16 will create:

de3a72c4505543b4be/d1f9bb1395b52a/de3a72c4-5055-43b4-bed1-f9bb1395c52a

This allows to inject measures and maybe process them in a single tool. This is
useful to generate load to test metricd and also profile the processing code.

See https://github.com/Pylons/webob/issues/355

The filter= argument is missing from the aggregation API.

This change adds it.

This change allow to pass the archive policy/unit when we batch
measures for metrics.

This change adds rateofchange method to Dynamic Aggregate API.

The current truncation code in AggregatedTimeSerie truncate on object
initialize only. That means that split up to POINTS_PER_SPLIT points are
stored, and that data is read and _truncated_ at this point.

When new points are added (existing merge), it's possible that the number of
points grows more than the number of points defined by the archive policy,
storing too many points. That's hidden because when re-read, the
AggregatedTimeSerie is truncated again.

However, that makes the storage driver store too much data.

This patches remove the max_size attribute from carbonara.AggregatedTimeSerie
and simply expose a truncate() method that can be used to truncate each split
based on the oldest point that is defined by the archive policy.

This change adds a new type of attribute for resource: datetime.

This allows to disable eager processing of new measures.

Fixes #507

This provides a new InfluxDB-compatible endpoint to write data to Gnocchi at
`/v1/influxdb'.

This design have some issue when two timeseries have the same names,
only the last one is dumped. Or if the same metric is asked
twice for whatever reason, measures have all timestamps duplicated.

This change the result to a new format:

For resources:
{
  "measures": { resource_id: {metric_name: {aggregation: MEASURES}}},
  "references": [RESOURCES_DETAILS]
}

For metrics
{
  "measures": {metric_id: {aggregation: MEASURES}},
  "references": [METRICS_DETAILS]
}

When the result is aggregtated we got:

For resources:
{
  "measures": { "aggregated": MEASURES},
  "references": [RESOURCES_DETAILS]
}
For metrics
{
  "measures": { "aggregated": MEASURES},
  "references": [METRICS_DETAILS]
}

Closes-bug: #477

dropna is like null but nan values are not returned.

This is the default for /v1/aggregates API.

This change allows to search for metrics with the same format as
resource search and cross metric aggregation.

And then it applies operations.

It also re-add transformation tests related to cross metric aggregation
endpoint but with new aggregates API.

Related #419

This changes implements the following operations for the aggregates API:
- rolling <agg> <window>
- resample <agg> <granularity>
- math (*/+-%...)
- abosolute/negative

It also re-add transformation tests related to metrics endpoint
but with new aggregates API.

Related #419

The option storage.aggregation_workers_number is not only used by storage, but
also by API cross aggregation. Make it generic and move the map_in_thread
method in a generic parallel_map function in gnocchi.utils.

This reverts commit 41117a83.

This feature will be implemented as part of
https://github.com/gnocchixyz/gnocchi/pull/419

The code is obsoleted by the future aggregates API

Related #419

Closes #374

this implements rolling aggregates which works similar to how
pandas rolling works but more trivial. this differs from deprecated
moving_average by correctly matching points to trailing timestamp
rather than leading timestamp.

it does not return the NaN values that result in shift in values.

Closes: #186

This change implements pagination links according the RFC5988.

It only implements the 'next' link, since other need extra mysql query
that we don't really want for now.

Related #3

This change adds an API to transform retrieved measurements.

First method implemented are absolute and negative. This API aims
to do some transformations that are not aggregation related.

This also fixes the data type passed to aggregated(), to ensure it's a
list and not an iterator.

if no boundary is set, the returned series does not necessarily
honour the overlap value. change behaviour so we require a boundary
if overlap is not 100%.

Fixes: #17

This change allows to compute the rate of change before applying
the aggregation method. To do so, aggregation method need to be
prefixed by "rate:"

As the wise Mehdi once said: "nobody reads documentation".
Everybody wants to run a daemon for the API server and everybody wants to run
`gnocchi-api'. Then let's provide it as a wrapper around uwsgi, so it's a fast
and performant HTTP by default.

deprecating moving average because the existing rolling average
for some reason generates a left-aligned rolling average which
is opposite of standard rolling average which should lag. also,
none of the existing parameters are configurable or are valid
(min_size is definitely wrong).

will improve this by leveraging pandas rolling

TEMP FIX:

don't unnecessarily pull in all granularities, search through
all series for min granularity, and rebuild.

just detect if valid, pull in granularity that's valid, and continue.

Sem-Ver: api-break

support ability to get multiple metrics based on id. returns result
with format matching:

{'metric_id1': [[timestamp, granularity, value]...],
 'metric_id2': [[timestamp, granularity, value]...]}

Fixes: #87

Authentication can be delegated to a frontend webserver/middleware, that
will ensure that REMOTE_USER env is set.

Change-Id: I8ebda5fccb0804d3349b84cefbb8353226436c19

It feels weird to have different option name for the same thing.

We don't need to provide the app.wsgi script anymore.
pbr provides it within the gnocchi-api binary

Change-Id: I6e16607128849a18b9e6eb1bc5558d1c9df64775

Change-Id: I0c0f190f556b36579a3afea63dc79fa7b1a9f443

This was a work around for old buggy Python 2.7 versions. Everybody will have
to upgrade!

Change-Id: Id105e2e8c1fab7af49c11b323db64886e57eade6