Filter widget based on datapoint values
Hi,
I suppose following on from my other message about combining CPU and memory in the same table widget, can these widgets be filtered - not dynamic/on demand but statically?
Take this example - can I have another table that only shows, for example, resources whose CPU OR Memory OR Uptime is beyond a certain threshold, e.g. ‘where CPU > X OR Memory > Y OR Uptime > Z’
Then expanding on that, I’m aiming for a dedicated dashboard highlighting ‘at risk’ or ‘concerning’ resources, so ideally this filter could be done at the dashboard level (e.g. with a ‘filter’ token) resulting in widgets on that dashboard inheriting the filter.
Thanks
ldoodle
Posted 2 years ago·Last reply 2 years ago
13 comments
LM User
·2 years agoYou’re getting ahead of yourself. You need to have the data before you can display it on the dashboard. So, you need a custom datasource where the instances of the datasources are the customer groups. The collection would collect the number of open, ack, sdt, etc. alerts per customer (i.e. per customer group). Once you have that, you’d put as many big number widgets or pie graphs as you want, filling in with the data from your datasource. You’d need to use an additional token that would match on the instance name, which would be the customer name.
ldoodle
OP2 years agoNo, not by group.
By “group” here, I mean each customer’s ‘Summary’ dashboard would have this Big Number widget with the breakdown, or even distinct Big Number widgets for each severity. Or even a Pie widget.
So the dashboard would have the ##defaultResourceGroup## token which would then inherit through to whichever widget I choose.
That way, would I still need to call the API for a particular group’s alerts? Or could I just return them all and then the dashboard token does the filtering?
LM User
·2 years agoNo, not by group. The portal stats has total alert counts. You could pretty easily build a datasource that hits the LM API to get the alert counts by group though. Your groups would be your instances and you’d call /device/groups/{groupID}/alerts to get the list of alerts that you can then count by severity.
ldoodle
OP2 years agoAh so you’re using a datasource which is either native or one you created. I was looking at doing something like an alerts Big Number widget to show a sum of number of critical, error and warning alerts for a given resource group.
Doesn’t seem like this capability is OOB.
LM User
·2 years agoThe left two are single datapoints on a single resource. The right two are single datapoints each from their own multi-instance datasources. First you add them in the “datapoints” section of the widget, in my case, i use the aggregate: sum method to dictate how the multiple numbers are aggregated together. Then in the numbers section, i tick the box “Change Number color according to thresholds” option.
ldoodle
OP2 years agoOh nice. How do you do that?
I mean I know to colourise them, but how do you do those queries?
LM User
·2 years agoI mean, I have a big number widget like this:
And I only pay attention to it if there is color.
ldoodle
OP2 years agoThen i have big number and table widgets with colorization that matches the thresholds
Do you mean you have Big Number that shows number of Critical alerts for example? How are you doing this? I’ve played with some datasources for this by when I click Save, I get ‘DataSource not found’ error.
ldoodle
OP2 years agoAs an aside, I’ve been playing with the NOC widget to try and fabricate something for now, except you can’t exclude ‘no alerts / green’ things.
I could have then had distinct widgets for Critical, Error and Warning. Is there any way around this?
ldoodle
OP2 years agoSince a DataSource (Datapoint) must return a numerical value, add the ability to filter based on the result per widget (any widget type) or per dashboard.
Something like this… in this Table widget example each column can has its own filter.
Optional: have 2 options, one for a ‘simple’ filter that uses the same mechanism as the color bar method (click operator buttons with a single value) and one for a complex filter that allows free text logic (like shown in the image), with the ability to use OR and AND operators etc.
LM User
·2 years agoCome up with text that others can copy/paste into the feature request system so we can get more traction behind useful features like this.
ldoodle
OP2 years agoYeah I thought that might be the case.
Thanks
LM User
·2 years agoNo, there’s no way to do that. The recommended way to accomplish what you’re trying to do is to set thresholds and use an alert list view.
I have a similar dashboard. First and foremost are a couple of alert list widgets that show the current alerts (they filters are too complex to have in one widget (LM, please come out with alert filtering language like LM Logs)). Then i have big number and table widgets with colorization that matches the thresholds. So i can look at the dashboard and quickly see if there are any alerts and if anything is colored yellow/orange/red. If so, something needs attention. I then put more detailed widgets below the fold so all i have to scroll down to see the details of what is going on.