Left-side derivative spikes

[cxreg]

Possibly related to #5806, if I am running a cross-series derivative, there are periodic spikes in the first position of the series. I have observed that these seem to coincide with the sample frequency of the series, so for example if series A records a sample 15 seconds before series B, then in that window between the two is when I will experience the spike.

Perhaps each series should look further back than it does? Maybe this is related to the fill() behavior in 5806?

[jsternberg]

This is likely related to #5733. We are going to be looking at derivative to fix this issue in 0.12 because the complexity for fixing this is likely not going to make it into the feature freeze for 0.11.

So that we can consider your use case when we start to try fixing this, what query are you doing, how, and what output you would expect?

Thank you for taking the time to file this bug.

[cxreg]

Yes it does look like the same issue, thanks!

[ariens]

Hi Folks, I'm exploring which TSDB to use for a new POC monitoring solution and have hit this bug. Are there are known work-arounds in v0.11 that I could use before v0.12 comes out? Are there any risks to this fix not being included in v0.12? Lastly--I see there's been some excellent progress on the completion of the milestone, is there an expected date of when a beta v0.12 will be ready?

Thanks!

[jsternberg]

A fix for aggregate queries and derivatives is in master now. Can you see if it works for your use case?

Thanks. More progress still needs to be made on this subject, but hopefully, it provides some relief.

[ariens]

Thanks, @jsternberg I've forwarded to someone whom might be able to provide feedback. I'm looking at OpenTSDB now that the clustering capabilities are no longer included in the open source versions.

[jsternberg]

I think this is resolved so I'm going to close this. #5943 will track this for other types of queries.

If there's any problem, please reopen. Thanks!