Commit Graph

3158 Commits

Author SHA1 Message Date
many
30a094cbb2
Change lacking errors 2021-11-03 14:33:33 +01:00
bors[bot]
c32f13a909
Merge #1800
1800: Analytics r=irevoire a=irevoire

Closes #1784
Implements [this spec](https://github.com/meilisearch/specifications/blob/update-analytics-specs/text/0034-telemetry-policies.md) 

# Anonymous Analytics Policy

## 1. Functional Specification

### I. Summary

This specification describes an exhaustive list of anonymous metrics collected by the MeiliSearch binary. It also describes the tools we use for this collection and how we identify a Meilisearch instance.

### II. Motivation

At MeiliSearch, our vision is to provide an easy-to-use search solution that meets the essential needs of our users. At all times, we strive to understand our users better and meet their expectations in the best possible way.

Although we can gather needs and understand our users through several channels such as Github, Slack, surveys, interviews or roadmap votes, we realize that this is not enough to have a complete view of MeiliSearch usage and features adoption. By cross-referencing our product discovery phases with aggregated quantitative data, we want to make the product much better than what it is today. Our decision-making will be taken a step further to make a product that users love.

### III. Explanation

#### General Data Protection Regulation (GDPR)

The metrics collected are non-sensitive, non-personal and do not identify an individual or a group of individuals using MeiliSearch. The data collected is secured and anonymized. We do not collect any data from the values stored in the documents.

We, the MeiliSearch team, provide an email address so that users can request the removal of their data: privacy@meilisearch.com.<br>
Thanks to the unique identifier generated for their MeiliSearch installation (`Instance uuid` when launching MeiliSearch), we can remove the corresponding data from all the tools we describe below. Any questions regarding the management of the data collected can be sent to the email address as well.

#### Tools

##### Segment

The collected data is sent to [Segment](https://segment.com/). Segment is a platform for data collection and provides data management tools.

##### Amplitude

[Amplitude](https://amplitude.com/) is a tool for graphing and highlighting collected data. Segment feeds Amplitude so that we can build visualizations according to our needs.

-----------
# The `identify` call we send every hour:

## System Configuration `system`

This property allows us to gather essential information to better understand on which type of machine MeiliSearch is used. This allows us to better advise users on the machines to choose according to their data volume and their use-cases.

 - [x] `system` => Never changes but still sent every hours
     - [x] distribution | On which distribution MeiliSearch is launched, eg: Arch Linux
     - [x] kernel_version | On which kernel version MeiliSearch is launched, eg: 5.14.10-arch1-1
     - [x] cores | How many cores does the machine have, eg: 24
     - [x] ram_size | Total capacity of the machine's ram. Expressed in `Kb`, eg: 33604210
     - [x] disk_size | Total capacity of the biggest disk. Expressed in `Kb`, eg: 336042103
     - [x] server_provider | Users can tell us on which provider MeiliSearch is hosted by filling the `MEILI_SERVER_PROVIDER` env var. This is also filled by our providers deploy scripts. e.g. GCP [cloud-config.yaml](56a7c2630c/scripts/providers/gcp/cloud-config.yaml (L33)), eg: gcp

## MeiliSearch Configuration

- [x] `context.app.version`: MeiliSearch version, eg: 0.23.0
- [x] `env`: `production` / `development`, eg: `production`
- [x] `has_snapshot`: Does the MeiliSearch instance has snapshot activated, eg: `true`

## MeiliSearch Statistics `stats`

 - [x] `stats`
     - [x] `database_size`: Size of indexed data. Expressed in `Kb`, eg: 180230
     - [x] `indexes_number`: Number of indexes, eg: 2
     - [x] `documents_number`: Number of indexed documents, eg: 165847
     - [x] `start_since_days`: How many days ago was the instance launched?, eg: 328

---------

- [x] Launched | This is the first event sent to mark that MeiliSearch is launched a first time

---------

- [x] `Documents Searched POST`: The Documents Searched event is sent once an hour. The event's properties are averaged over all search operations during that time so as not to track everything and generate unnecessary noise.
  - [x] `user-agent`: Represents all the user-agents encountered on this endpoint during one hour, eg: `["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]`
  - [x] `requests`
      - [x] `99th_response_time`: The maximum latency, in ms, for the fastest 99% of requests, eg: `57ms`
      - [x] `total_suceeded`: The total number of succeeded search requests, eg: `3456`
      - [x] `total_failed`: The total number of failed search requests, eg: `24`
      - [x] `total_received`: The total number of received search requests, eg: `3480`
  - [x] `sort`
      - [x] `with_geoPoint`: Does the built-in sort rule _geoPoint rule has been used?, eg: `true` /`false`
      - [x] `avg_criteria_number`: The average number of sort criteria among all the requests containing the sort parameter. "sort": [] equals to 0 while not sending sort does not influence the average, eg: `2`
  - [x] `filter`
      - [x] `with_geoRadius`: Does the built-in filter rule _geoRadius has been used?, eg: `true` /`false`
      - [x] `avg_criteria_number`: The average number of filter criteria among all the requests containing the filter parameter. "filter": [] equals to 0 while not sending filter does not influence the average, eg: `4`
      - [x] `most_used_syntax`: The most used filter syntax among all the requests containing the requests containing the filter parameter. `string` / `array` / `mixed`, `mixed`
  - [x] `q`
      - [x] `avg_terms_number`: The average number of terms for the `q` parameter among all requests, eg: `5`
  - [x] `pagination`:
      - [x] `max_limit`: The maximum limit encountered among all requests, eg: `20`
      - [x] `max_offset`: The maxium offset encountered among all requests, eg: `1000` 

---

- [x] `Documents Searched GET`: The Documents Searched event is sent once an hour. The event's properties are averaged over all search operations during that time so as not to track everything and generate unnecessary noise.
  - [x] `user-agent`: Represents all the user-agents encountered on this endpoint during one hour, eg: `["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]`
  - [x] `requests`
      - [x] `99th_response_time`: The maximum latency, in ms, for the fastest 99% of requests, eg: `57ms`
      - [x] `total_suceeded`: The total number of succeeded search requests, eg: `3456`
      - [x] `total_failed`: The total number of failed search requests, eg: `24`
      - [x] `total_received`: The total number of received search requests, eg: `3480`
  - [x] `sort`
      - [x] `with_geoPoint`: Does the built-in sort rule _geoPoint rule has been used?, eg: `true` /`false`
      - [x] `avg_criteria_number`: The average number of sort criteria among all the requests containing the sort parameter. "sort": [] equals to 0 while not sending sort does not influence the average, eg: `2`
  - [x] `filter`
      - [x] `with_geoRadius`: Does the built-in filter rule _geoRadius has been used?, eg: `true` /`false`
      - [x] `avg_criteria_number`: The average number of filter criteria among all the requests containing the filter parameter. "filter": [] equals to 0 while not sending filter does not influence the average, eg: `4`
      - [x] `most_used_syntax`: The most used filter syntax among all the requests containing the requests containing the filter parameter. `string` / `array` / `mixed`, `mixed`
  - [x] `q`
      - [x] `avg_terms_number`: The average number of terms for the `q` parameter among all requests, eg: `5`
  - [x] `pagination`:
      - [x] `max_limit`: The maximum limit encountered among all requests, eg: `20`
      - [x] `max_offset`: The maxium offset encountered among all requests, eg: `1000` 

---

- [x] `Index Created`
  - [x] `user-agent`: Represents the user-agent encountered for this API call, eg: ["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]
  - [x] `primary_key`: The name of the field used as primary key if set, otherwise `null`, eg: `id`

---

- [x] `Index Updated`
  - [x] `user-agent`: Represents the user-agent encountered for this API call, eg: ["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]
  - [x] `primary_key`: The name of the field used as primary key if set, otherwise `null`, eg: `id`

---

- [x] `Documents Added`: The Documents Added event is sent once an hour. The event's properties are averaged over all POST /documents additions operations during that time to not track everything and generate unnecessary noise.
  - [x] `user-agent`: Represents the user-agent encountered for this API call, eg: ["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]
  - [x] `payload_type`: Represents all the `payload_type` encountered on this endpoint during one hour, eg: [`text/csv`]
  - [x] `primary_key`: The name of the field used as primary key if set, otherwise `null`, eg: `id`
  - [x] `index_creation`: Does an index creation happened, eg: `false`

---

- [x] `Documents Updated`: The Documents Added event is sent once an hour. The event's properties are averaged over all PUT /documents additions operations during that time to not track everything and generate unnecessary noise.
  - [x] `user-agent`: Represents the user-agent encountered for this API call, eg: ["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]
  - [x] `payload_type`: Represents all the `payload_type` encountered on this endpoint during one hour, eg: [`application/json`]
  - [x] `primary_key`: The name of the field used as primary key if set, otherwise `null`, eg: `id`
  - [x] `index_creation`: Does an index creation happened, eg: `false`

---

- [x] Settings Updated
  - [x] `user-agent`: Represents the user-agent encountered for this API call, eg: ["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]
  - [x] `ranking_rules`
      - [x] `sort_position`: Position of the `sort` ranking rule if any, otherwise `null`, eg: `5`
  - [x] `sortable_attributes`
      - [x] `total`: Number of sortable attributes, eg: `3`
      - [x] `has_geo`: Indicate if `_geo` is set as a sortable attribute, eg: `false`
  - [x] `filterable_attributes`
      - [x] `total`: Number of filterable attributes, eg: `3`
      - [x] `has_geo`: Indicate if `_geo` is set as a filterable attribute, eg: `false`

---

- [x] `RankingRules Updated`
  - [x] `user-agent`: Represents the user-agent encountered for this API call, eg: ["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]
  - [x] `sort_position`: Position of the `sort` ranking rule if any, otherwise `null`, eg: `5`

---

- [x] `SortableAttributes Updated`
  - [x] `user-agent`: Represents the user-agent encountered for this API call, eg: ["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]
  - [x] `total`: Number of sortable attributes, eg: `3`
  - [x] `has_geo`: Indicate if `_geo` is set as a sortable attribute, eg: `false`

---

- [x] `FilterableAttributes Updated`
  - [x] `user-agent`: Represents the user-agent encountered for this API call, eg: ["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]
  - [x] `total`: Number of filterable attributes, eg: `3`
  - [x] `has_geo`: Indicate if `_geo` is set as a filterable attribute, eg: `false`

---

- [x] Dump Created
  - [x] `user-agent`: Represents the user-agent encountered for this API call, eg: ["MeiliSearch Ruby (2.1)", "Ruby (3.0)"]

---

Ensure the user-id file is well saved and loaded with:
- [x] the dumps
- [x]  the snapshots



- [x] Ensure the CLI uuid only show if analytics are activate at launch **or already exists** (=even if meilisearch was launched without analytics)

Co-authored-by: Tamo <tamo@meilisearch.com>
Co-authored-by: Irevoire <tamo@meilisearch.com>
2021-10-29 16:11:03 +00:00
marin postma
519093ea65
fix bad rebase 2021-10-29 17:32:49 +02:00
Tamo
bd49d1c4b5
fix one small bug 2021-10-29 17:25:56 +02:00
marin postma
2665c0099d
clippy + fmt 2021-10-29 17:25:56 +02:00
marin postma
d65f055030
pass anaytics into Arc instead of static ref 2021-10-29 17:25:55 +02:00
Tamo
66d87761b7
align the parameters in the launche resume 2021-10-29 17:25:55 +02:00
Tamo
ba69ad672a
fix the timing issue 2021-10-29 17:25:55 +02:00
Tamo
7934e3956b
replace all mutexes by channel 2021-10-29 17:25:55 +02:00
Guillaume Mourier
68fe93b7db
add ranking_rules marker before sort_position 2021-10-29 17:25:55 +02:00
Tamo
efd0ea9e1e
makes clippy happier 2021-10-29 17:25:55 +02:00
Tamo
6ef73eb226
fix all the single settings route and add the searchable attributes Updated event 2021-10-29 17:25:55 +02:00
Tamo
fc2f23d36c
move the start_since_days to teh root of the identify 2021-10-29 17:25:54 +02:00
Tamo
7c39fab453
move the user-agent out of the context in every request 2021-10-29 17:25:54 +02:00
Tamo
c5164c01c0
set the total of sortable attributes and filterable-attributes to 0 when not set 2021-10-29 17:25:54 +02:00
Tamo
351ad32d77
fix the index_creation boolean 2021-10-29 17:25:54 +02:00
Tamo
3ad8311bdd
split the analytics in a module 2021-10-29 17:25:54 +02:00
Tamo
ea5ae2bae5
sort the imports 2021-10-29 17:25:54 +02:00
Tamo
72e3adc55e
display an instance-id instead of a user-id 2021-10-29 17:25:54 +02:00
Tamo
b250392e8d
remove the first - in the path to the db instance in the instance-id 2021-10-29 17:25:53 +02:00
Tamo
d8b0d68840
use a regex to count the number of filters instead of split + flatten 2021-10-29 17:25:53 +02:00
Tamo
c4737749ab
bump segment to be able to display a user 2021-10-29 17:25:53 +02:00
Tamo
a1ab02f9fb
remove some commented code 2021-10-29 17:25:53 +02:00
Tamo
bba64b32ca
async_traits is not needed anymore 2021-10-29 17:25:53 +02:00
Tamo
9abd2aa9d7
make the analytics interval a const 2021-10-29 17:25:53 +02:00
Tamo
de35a9a605
use an official release of segment 2021-10-29 17:25:53 +02:00
Tamo
ed750e8792
fix start_since_day 2021-10-29 17:25:53 +02:00
Tamo
37ca50832c
fix the sort position 2021-10-29 17:25:52 +02:00
Tamo
31c7a0105b
fix a bug on the batch documents function 2021-10-29 17:25:52 +02:00
Tamo
ddab9eafa1
fix a typo 2021-10-29 17:25:52 +02:00
Tamo
76a4f86e0c
rename user-id to instance-uid 2021-10-29 17:25:52 +02:00
Tamo
6b34318274
makes clippy happy 2021-10-29 17:25:52 +02:00
Tamo
5508c6c154
a bit of styling 2021-10-29 17:25:52 +02:00
Tamo
9a62ac0c94
send the analytics only once every hours 2021-10-29 17:25:52 +02:00
Tamo
01737ef847
remove all the debug prints 2021-10-29 17:25:51 +02:00
Tamo
3144b572c4
remove the debug mode in release 2021-10-29 17:25:51 +02:00
Tamo
10de92987a
compile write_user_id only when the analytics are enabled 2021-10-29 17:25:51 +02:00
Tamo
c752c14c46
refactorize the dump and snapshot 2021-10-29 17:25:51 +02:00
Tamo
87a8bf5e96
write and load the user-id in the dumps 2021-10-29 17:25:51 +02:00
Tamo
ba14ea1243
plug the new batchers into the documents route 2021-10-29 17:25:51 +02:00
Tamo
9be90011c6
save the user-id in the config dir of the OS 2021-10-29 17:25:51 +02:00
Tamo
f9b14ca149
simplify the search batcher 2021-10-29 17:25:50 +02:00
Tamo
6591acfdfa
rename the documents batchers 2021-10-29 17:25:50 +02:00
Tamo
e64ba122e1
factorize the code between the two documents batcher 2021-10-29 17:25:50 +02:00
Tamo
a9523146a3
simplify the into_events methods 2021-10-29 17:25:50 +02:00
Tamo
392ee86714
implement the documents batcher 2021-10-29 17:25:50 +02:00
Tamo
1d73f484f0
update the primary key when creating a new index 2021-10-29 17:25:50 +02:00
Tamo
cfcd3ae048
move the version to context.app 2021-10-29 17:25:50 +02:00
Tamo
5395041dcb
fix the stats and stop sending events when no request happened 2021-10-29 17:25:49 +02:00
Tamo
40eabd50d1
integrate the search batcher in the search route 2021-10-29 17:25:49 +02:00