Chart data consistency


In some of my dashboard charts I am seeing very big differences in the height of peaks for past periods when I simply refresh the page.
I was wondering if there is any logical explanation for this, or whether it is maybe a bug?

For example, I have the following query linked to a chart:
SELECT average(duration * 1000) AS ‘Response time’ FROM Transaction WHERE transactionType = ‘Web’ AND request.uri LIKE ‘/example/webservice/searchItems’ TIMESERIES SINCE 3600 seconds ago EXTRAPOLATE

Peaks in this chart often vary massively in height when I refresh the page, while keeping the settings/time window the same.
I understand that a peak for the “current time”/datapoint can get higher or lower since new data is still coming in, affecting the average.
Also when you change the time frame for the chart, the resolution changes, so the plotted average might then be for a greater or smaller amount of time.
The same goes for when you change the width the chart is displayed at, the x-axis changes, possibly affecting the averages in each time period.

However, when the query, the time window and the width of the chart all stay the same: how would the average for a data point half an hour in the past fluctuate greatly?

Even when I go to the query builder and use the same query from above and click ‘Run’, I get one version of the chart.
I click ‘Run’ again 1 second later, and get a very different version of the chart. More specifically it seems to alternate between 2 versions of the chart (mainly seen by the height of the peak/the numbers on the y-axis). Sometimes I have to click run 10 times to get the ‘different version’, sometimes it only takes 1 or 2 clicks.
One version of the chart is clearly ‘dominant’ in the sense that I see that one 90% of the time but every few refreshes/runs I will see the other one.

The “dominant” version of the chart, generated around 13:00PM, with a peak of around 95 around 12:20PM:

The other version of the chart which I see every few runs/refreshes, generated a few seconds later, for the same query, with a peak of around 420 around 12:20PM:

In case it matters: I have tried the query without EXTRAPOLATE as well, but this makes no difference/exhibits the same issue (I don’t think there’s enough data coming in yet for there to be any uncollected samples).

Currently I don’t know which version is correct/the “truth” and why it alternates between refreshes.

Does anybody have an idea why it would be showing this behavior?



Hi @newrelic-jeroen - Have you tried adding LIMIT MAX to the end of the query and removing the EXTRAPOLATE?

Hi @stefan_garnham - Thanks for your reply!
I had not tried that yet, but just did and with a LIMIT MAX the problem unfortunately persists.

While testing this I noticed that if I specify an exact time period, with for example:

TIMESERIES SINCE ‘2021-11-09 12:00:00+0100’ UNTIL ‘2021-11-09 13:00:00+0100’
(which corresponds with the period in the charts in the attached images, but was then ‘3600 seconds ago’)

The problem does not occur, regardless of with or without EXTRAPOLATE or LIMIT MAX.

However, when I use a period that includes “now”, for example:

TIMESERIES SINCE 3600 seconds ago

Then the chart behaves as described in the opening post.

Ok, perhaps someone from support will be able to provide a better understanding.

One would think/hope…

Hey Jeroen,

I just reached out to you via support ticket to look into this further. Please let me know :slight_smile:

Hey Pete,

The support ticket doesn’t work anymore apparently. Never got a confirmation of a reply I sent Friday, and now when I visit the ticket URL, I get:

URL No Longer Exists
You have attempted to reach a URL that no longer exists on

You may have reached this page after clicking on a direct link into the application. This direct link might be:
• A bookmark to a particular page, such as a report or view
• A link to a particular page in the Custom Links section of your Home Tab, or a Custom Link
• A link to a particular page in your email templates

If you reached this page through a bookmark, you are probably trying to access something that has moved. Please update your bookmark.

If you reached this page through any of the other direct links listed above, please notify your administrator to update the link.

Thank you again for your patience and assistance. And thanks for using!