bug in arithmetic metaqueries

[gwuillou]

I am doing a sum of 3 queries A + B +C . I get the follow result:

But if I zoom out, there is not anymore this 0 value in the left corner and everything works.

I am using influxdb and grafana 5.1.3.

[Gauravshah]

@gwuillou when zoomed-in are the results correct ? When zoomed out are able to see A,B,C ( without applying A+B+C Query) ?

[gwuillou]

Here different zooms. There is not always the problem apparently!

[Gauravshah]

can you send a screenshot of your panel as well ( dashboard config )

[gwuillou]

issues_MQ7.pdf

[Gauravshah]

I think the problem is because metaqueries does not assume null as 0. So for the sections where 1 of the (a,b or c ) value is null then the resultant D is null. You can probably try (((a)? a :0) + ((b)? b :0) + ((c)? c :0))

[gwuillou]

Hello, I have tried your proposition without success: (((A['siesa'])?A['siesa']:0)+((C['djeva'])?C['djeva']:0)+((B['sinergy'])?B['sinergy']:0)) I don’t know if you are other idea?

Thank for your help,

Gaëtan

Le 5 juil. 2018 à 13:19, Gaurav M Shah notifications@github.com a écrit :

I think the problem is because metaqueries does not assume null as 0. So for the sections where 1 of the (a,b or c ) value is null then the resultant D is null. You can probably try (((a)? a :0) + ((b)? b :0) + ((c)? c :0))

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/GoshPosh/grafana-meta-queries/issues/29#issuecomment-402690471, or mute the thread https://github.com/notifications/unsubscribe-auth/ATtI571lm1vUwgNc5ricvgPOMDFG2B09ks5uDfYtgaJpZM4VDYl2.

[Gauravshah]

I am not sure, I am unable to reproduce the error on my end. Looks to me that one of the datapoint is throwing the graph off. Can you try plotting a table instead and look at the odd data point ?

[gwuillou]

When I am exporting the csv:

Conso_FMV;2018-07-05T21:00:00+02:00;30.117048 Conso_FMV;2018-07-05T21:15:00+02:00;29.324952 Conso_FMV;2018-07-05T21:30:00+02:00;29.627704 Conso_FMV;2018-07-05T21:45:00+02:00;29.00208 Conso_FMV;2018-07-05T22:00:00+02:00;29.310443999999997 Conso_FMV;2018-07-05T22:15:00+02:00;29.481824 Conso_FMV;2018-07-05T22:30:00+02:00;28.184124 Conso_FMV;2018-07-05T22:45:00+02:00;28.682604 Conso_FMV;2018-07-05T23:00:00+02:00;28.003059999999998 Conso_FMV;2018-07-05T23:15:00+02:00;26.963816 Conso_FMV;2018-07-05T23:30:00+02:00;26.033584000000005 Conso_FMV;2018-07-05T23:45:00+02:00;26.090695999999998 Conso_FMV;2018-07-06T00:00:00+02:00;24.468472 Conso_FMV;2018-06-25T02:00:00+02:00;0 Conso_FMV;2018-07-06T00:15:00+02:00;0 Conso_FMV;2018-07-06T00:30:00+02:00;0 Conso_FMV;2018-07-06T00:45:00+02:00;0 Conso_FMV;2018-07-06T01:00:00+02:00;0 Conso_FMV;2018-07-06T01:15:00+02:00;0 Conso_FMV;2018-07-06T01:30:00+02:00;0 Conso_FMV;2018-07-06T01:45:00+02:00;0 Conso_FMV;2018-07-06T02:00:00+02:00;0 Conso_FMV;2018-07-06T02:15:00+02:00;0 Conso_FMV;2018-07-06T02:30:00+02:00;0

Le 9 juil. 2018 à 11:28, Gaurav M Shah notifications@github.com a écrit :

I am not sure, I am unable to reproduce the error on my end. Looks to me that one of the datapoint is throwing the graph off. Can you try plotting a table instead and look at the odd data point ?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/GoshPosh/grafana-meta-queries/issues/29#issuecomment-403418030, or mute the thread https://github.com/notifications/unsubscribe-auth/ATtI55md7cMzroWNSsLp9b6K2khbG3oBks5uEyJSgaJpZM4VDYl2.

[Gauravshah]

can you give a try to (((A['siesa'])?Math.round(A['siesa']*100)/100:0)+((C['djeva'])?Math.round(C['djeva']*100)/100:0)+((B['sinergy'])?Math.round(B['sinergy']*100)/100:0))

[gwuillou]

Always the same problem!

Le 9 juil. 2018 à 11:39, Gaurav M Shah notifications@github.com a écrit :

can you give a try to (((A['siesa'])?Math.round(A['siesa']100)/100:0)+((C['djeva'])?Math.round(C['djeva']100)/100:0)+((B['sinergy'])?Math.round(B['sinergy']*100)/100:0))

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/GoshPosh/grafana-meta-queries/issues/29#issuecomment-403420994, or mute the thread https://github.com/notifications/unsubscribe-auth/ATtI5_BBN9cnzWG1VrMl6AaNC_xQLdmJks5uEyTigaJpZM4VDYl2.

[Gauravshah]

its hard to debug without that dataset

[gwuillou]

Perhaps it can help you:

Le 9 juil. 2018 à 11:44, Gaurav M Shah notifications@github.com a écrit :

its hard to debug without that dataset

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/GoshPosh/grafana-meta-queries/issues/29#issuecomment-403422232, or mute the thread https://github.com/notifications/unsubscribe-auth/ATtI5_nEi4SH7LOHi6Dg3TIRmeneIb2Fks5uEyXggaJpZM4VDYl2.

[DerAndereAndi]

I have similar drawing issues. I can reproduce it when showing 2 datasets with data available in different time intervals and selecting an group by time() interval that is smaller than the first but larger than the second.

E.g. Dataset 1 is available in 10 Minutes interval, dataset 2 in 5 Minutes intervals. using group by time(10m) shows nice graphs, but using group by time (5m) shows issues in the graph. I am not arguing about the sense of this grouping, but to provide an example on how to reproduce graphing issues with MetaQueries.

Hope this helps.

[Gauravshah]

@TheRealKerni yup that scenario makes sense. but I am unsure if we can default numbers to 0. Its more general grafana issue than this plugin itself. I think it behaves in the same way if a datasource itself would have had no points in that location

[DerAndereAndi]

I am not sure what is going on yet, tried to debug a bit but haven't find the cause.

But I have a small example animated gif that may help.

The data source has data that is fetched once a minute, the minimum interval for the chart is set to 5 minutes. When showing the influxDB data, everything is fine, when showing the MetaQuery instead it acts app. I found that the timestamp of the last item when using MetaQuery suddenly changes to the first timestamp in the dataset!

Take a look.

[Gauravshah]

@ShilpaSivanesan we should see if this problem exists in a generic fashion. Lets look into this one next week.

[tiago-cruz-movile]

Hello,

I have two timestamps, and I would like to see the difference between using B['dequeuedAt']-A['routedAt']

Sometimes I've got strange negative values, like this:

How can I avoid null values?

I've tried (((B['dequeuedAt'])? B['dequeuedAt'] :0) - ((A['routedAt'])? A['routedAt'] :0) and ((B['dequeuedAt'])?Math.round(B['dequeuedAt']*100)/100:0) - ((A['routedAt'])?Math.round(A['routedAt']*100)/100:0)

But without success.

Sometimes the count got really strange like this:

Thanks!!

[Gauravshah]

@tiago-cruz-movile you should put it on a table to see what results do you get. also you can try ( (B['dequeuedAt'] && A['routedAt'])? B['dequeuedAt'] - A['routedAt'] : 0 )

[aquilax]

I had similar problem two queries A and B and the B query did not return values for every time interval for which A did. What worked for me was using something like A['key1'] + (B ? B['key2'] : 0) which essentially returns 0 if the B query does not have a value.

[ZyanKLee]

today I noticed a very similar behaviour with a simple A-B logic (well, B-A really):

A: SELECT mean("value") AS "Free" FROM "node_memory_MemFree_bytes" WHERE $timeFilter GROUP BY time($__interval) fill(none) B: SELECT mean("value") AS "Total" FROM "node_memory_MemTotal_bytes" WHERE $timeFilter GROUP BY time($__interval) fill(none) C: B["Total"]-A["Free"]

At first I had the time() set to 10s statically - and when I plotted 13 days or more the effect would look like that:

then I went back to letting grafana determine the grouping interval and it solved the problem for me. (tested up till 30 days worth of data)

[cassiel74]

Can't get metaqueries to work like you do... Here's my setup, what I am missing?

Regards

[DerAndereAndi]

@cassiel74 add the value $col into the big ALIAS BY field.

The problem is that if ALIAS BY is not set, then the actual identifier prepended by values.. The $col value will eliminate this prepending string.

[cassiel74]

Ah... I put the alias in SQL style :-) Now it's working, using only ALIAS BY and nothing in SELECT statement

[Gauravshah]

thank you @TheRealKerni

[frams]

I was having the same issue I solve mine by grouping by time ($__interval, previous) on both A AN B queries (MYSQL Data Source) and that seems to have fixed the issue.