[monitoring] Adding influxDB 2.x version support #274 #584
Conversation
praptisharma28
force-pushed
the
influxdb2.x
branch
from
e2b2446 to
295446d
Compare
praptisharma28
force-pushed
the
influxdb2.x
branch
2 times, most recently
from
83694c5 to
cfe8da0
Compare
praptisharma28
force-pushed
the
influxdb2.x
branch
from
87f36b9 to
d02db9c
Compare
praptisharma28
force-pushed
the
influxdb2.x
branch
from
5647cd4 to
0aa3c6d
Compare
|
June 18th weekly call summary:
|
Fixes #274
There was a problem hiding this comment.
There's too many if statements in the code (eg: if influxdb1 or influxdb2).
This is not the way to go. The logic for influxdb1 or influxdb2 must be encapsulated in each respective timeseries DB backend, the rest of the code should just be adapted to call that logic.
Please have a look at the work which was done for the elasticsearch backend to get an idea:
#164
There was a problem hiding this comment.
In our call today, @praptisharma28 walked me through the code and helped me understand the changes. In this process, we found some areas for improvement.
| timezone=settings.TIME_ZONE | ||
| ): | ||
| bucket = self.bucket | ||
| measurement = params.get('measurement') |
There was a problem hiding this comment.
params.get('measurement') is equal to params.get('key'). Let's avoid duplicating the values that are already present to the method.
| 'start_date': start_date, | ||
| 'end_date': end_date, | ||
| 'measurement': self.config_dict.get('measurement', self.metric.key), | ||
| 'field_name': fields or self.config_dict.get('field_name'), |
There was a problem hiding this comment.
If you are not using the field_name in the query, then please remove it.
There was a problem hiding this comment.
I used pdb, got:
(Pdb) params
{'field_name': None, 'key': 'test_metric', 'time': '2024-06-25', 'days': '7d', 'content_type': 'openwisp_users.user', 'object_id': '4f5f77d3-3b30-4eca-97c6-1301217a4edc', 'start_date': None, 'end_date': None, 'measurement': 'test_metric'}
But if I remove it, I lose on the summary of uptime chart. Charts like mostly reachable, unreachable, partially reachable are lost.
There was a problem hiding this comment.
There should be only one summary in the uptime chart.
| params.update({ | ||
| 'start_date': start_date, | ||
| 'end_date': end_date, | ||
| 'measurement': self.config_dict.get('measurement', self.metric.key), |
There was a problem hiding this comment.
This is duplicate to key, let's remove it.
| return timeseries_db._get_top_fields( | ||
| query=q, | ||
| default_query=self._default_query, | ||
| query=self.get_query(), | ||
| chart_type=self.type, | ||
| group_map=self._get_group_map(params['days']), | ||
| number=number, |
There was a problem hiding this comment.
The InfluxDB 2 client does not accept default_query keyword argument. We need to ensure that this works at par with InfluxDB 1 and add any required tests for this method.
| points = summary = timeseries_db._get_top_fields( | ||
| default_query=self._default_query, | ||
| chart_type=self.type, | ||
| group_map=self.GROUP_MAP, | ||
| number=self.top_fields, | ||
| params=self._get_query_params(self.DEFAULT_TIME), | ||
| time=time, | ||
| query=self.query, | ||
| get_fields=False, |
There was a problem hiding this comment.
This change is wrong. Please revert them to what was here before.
| summary = timeseries_db.get_list_query(summary_query) | ||
| points = timeseries_db.get_list_query(data_query, key=self.metric.key) | ||
| summary = timeseries_db.get_list_query( | ||
| summary_query, key=self.metric.key |
There was a problem hiding this comment.
| summary_query, key=self.metric.key | |
| summary_query, |
I don't see the key argument being used in the timeseries_db.get_list_query method. Remove it if it is not needed.
| for point in points: | ||
| time_value = point.get('time') or point.get('_time') |
There was a problem hiding this comment.
Let's avoid doing this and rename the time field in the flux query.
| if not time_value: | ||
| logger.warning(f"Point missing time value: {point}") | ||
| continue |
There was a problem hiding this comment.
Will this code ever get executed? Isn't every point in timeseries db mapped to a timestamp.
| if decimal_places and isinstance(value, (int, float)): | ||
| if decimal_places is not None and value is not None: |
There was a problem hiding this comment.
Can you please verify the error you were getting here?
| def get_ping_data_query(self, bucket, start, stop, device_ids): | ||
| device_filter = ' or '.join([f'r["object_id"] == "{id}"' for id in device_ids]) | ||
| query = f''' | ||
| from(bucket: "{bucket}") | ||
| |> range(start: {start}, stop: {stop}) | ||
| |> filter(fn: (r) => r["_measurement"] == "ping") | ||
| |> filter(fn: (r) => r["_field"] == "loss" or r["_field"] == "reachable" or r["_field"] == "rtt_avg" or r["_field"] == "rtt_max" or r["_field"] == "rtt_min") | ||
| |> filter(fn: (r) => r["content_type"] == "config.device") | ||
| |> filter(fn: (r) => {device_filter}) | ||
| |> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false) | ||
| |> yield(name: "mean") | ||
| ''' | ||
| return query |
There was a problem hiding this comment.
Please move this query to queries.py and use get_query() method to generate the final query.
| bucket = self.bucket | ||
|
|
||
| # Start building the Flux query | ||
| flux_query = f'from(bucket:"{bucket}")' |
There was a problem hiding this comment.
@praptisharma28 why are you building the query like this here? Why aren't you using the queries defined in queries.py?
This looks wrong. Charts for different metrics are handled differently. That's why there are individual queries for each chart.
There was a problem hiding this comment.
My bad, I got confused between the get_list_queries and read method. The current implementation is okay.
Fixes #274
Checks: