[no value] in notifications

[mateocarusotti]

(Sorry for my english)
Hi, im using Grafana for Practice in my college. I create a few alerts, thats alerts check de usage memory of two honeypots (an alert per each honeypot). My problem is that frecuely i receive an alert in state Firing with [no value] inside of notification, and after that inever receive the notification with Resolved state.
Here i share my alerts and some configurations:

1. The queries and the Threshold:
   

   

2. No data and error handler
   

   

3. The summary where i got [no value]
   

   

4. Period Pending = 2m, Evaluation Interval= 2m

Any idea what the error could be?

[dawiddebowski]

Hi, can you share what the notification looks like?
Another question - if, just for tests, you’ll set Alert state if no data or all values are null (second screen) to No Data will you get another notification?

What I’m pointing at is [no value] seems like no data error - in this case it might happen when either of the metrics don’t have the data. Are you using regular node exporter? Also could you share the result of the queries in some screenshot (if the data aren’t a security threat).

[volbulbul]

I had the same problem for 2 days after upgrade to newer version of grafana tried to stop alerts silences but didnt solve it. I completely stoped grafana but I am still getting messages. I think the problem is releated with telegram webhooks but dont know how to reset bot ?

[mateocarusotti]

Hi! Thank for your response

1. This is the telegram notification i receive
   
   
   
   1.1) Here is my template just in case
   
   
2. Thank for the tip, today i set that option, I haven’t received anything yet but the notifications are not that frequent.
   
   
3. If im using regular node_exporte
   I think so, but i dont have access to that directly but i believe that is a normal node_exporter
4. Here are the result, is just the usage of ram
   
   

With my profesor think that could be a error of node_exporter(or other process) or the BD (he resently migrated from SQLlite to MySQL)

[dawiddebowski]

I can think of 2 scenarios:

1. Node exporter / prometheus does not keep up with data ingestion, hence sometimes the data is not there at the time of querying. You could mitigate that by getting some avg_over_time or using Range query with Reduce operation other than last or, if you do not want any of this - using offset 1m with your query to get the point that definitely should already be there. Another way is to (if it’s possible) migrate to VictoriaMetrics - it’s prometheus compatible and AFAIK if the value isn’t there it looks some time in the past to get the most recent value.
2. The node restarts / there are no series / there are scrape errors - this should be easy to check if prometheus scrapes itself - there should be some metric with scrape errors - if this is more than 0 the points might not have been scrape (Grafana will extrapolate the plot in the graphs with expressions, so it might not be visible).

[mateocarusotti]

Thank you very much for taking the time to answer my question. I recently changed the No Data handler to Normal, just to see what happens. If I see something interesting, I will share it.