#influxdb

phendryx in regards to my question the other day, about the CQ resampling, i'm looking for some documentation for RESAMPLE, i can only find what you've sent, and all it documents is how to do it for NOW - x, not from the last X entries or TIMESTAMP - x

do you know where i can find more documentation on that?

how to profile influxdb queries?

Danielss89 thats all the documentation you really need. There are 2 use cases for resampling.

1) when everything is flowing smoothly, you might get some data a little late. Having a CQ with RESAMPLE options will allow you to check the last x iterations. In this case, the following query RESAMPLES the last 2 groups of 30 minutes to ensure the data is correct in your rollup measurement:

CREATE CONTINUOUS QUERY vampires_1 ON transylvania RESAMPLE FOR 60m BEGIN SELECT count(dracula) INTO vampire_populations_1 FROM raw_vampires GROUP BY time(30m) END

2) when you need to backfill a rollup from "last week" or something along those lines. Thats when you use a query that looks something like:

SELECT min(temp) as min_temp, max(temp) as max_temp INTO "reading.minmax.5m" FROM reading WHERE time >= '2015-12-14 00:05:20' AND time < '2015-12-15 00:05:20' GROUP BY time(5m)

hello, I have a small question: is it possible to add new tag, when downsampling data? That tag doesn't exist in the table from which data is downsampled.

tommyboy113 are you downsampling to a new measurement?

yes

actually, that question was pointless, you can always add a new tag, sorry

my question, not yours :)

wait.. thats downsampling, not via http or api, etc..

I spoke too soon

phendryx ok, well my "problem" is that a device might send a datapoint from 1 year ago, so the resample would have to be from the period where the time is from

to be honest it doesn't matter. I'm trying to use it for data migration. I'm downsampling old data and I need to add new tags.

phendryx and as i understand it, a CQ uses NOW()?

and not the actual time from the datapoint

I can use whatever I want, api, http and so on.

tommyboy113 then if I were doing data migration, Id write a script in ruby/python/etc that would select my data out, allow me to transform it, and push it back into the new measurement

otherwise Id try to find a built in solution.. guess it depends on your data size too

Danielss89 thats the thing, CQ with RESAMPLE is not meant to catch data from a year ago

phendryx yeah thought so :( so how would i accomplish this? I mean, i don't know when a datapoint like this comes in

@phendryx okay, got it. So as I understand from the direction I should take is I can't add new tags in select into statements ant so on?

Danielss89 that's a good question... for work, I had to build a "replay" system that its purpose in life was backfilling data and then running downsampling/rollup

what about kapacitor, can that do what i want?

or would that be the same?

Danielss89 if you have 2 retention policies, 1 for 30 days of all data (as it flows in, not necessarily the timestamp of the record), another for all downsampled data, you could write some sort of check that looks for data older than 30 days and then do a manual downsample

Danielss89 thats interesting.. I havent looked into that before.. maybe?

tommyboy113 yeah, I dont think so.. https://github.com/influxdata/influxdb/issues/3904 shows other people wanting the same, but the issue was closed

phendryx, okay, thanks :)

you're only able to insert one 'row' per timestamp, correct? there is no way to augment it's uniqueness (tags, etc)?

*its