Hello everyone, I have faced an issue regarding Azure Database for PostgreSQL -flexible server metrics, which I hoped someone here can help me solve it. I have an Azure Database for PostgreSQL - flexible server, which has an alert configured on it, so I will be notified if the number of failed connections on the server exceeds a threshold, it uses the connections_failed metric, that is supposedly enabled by default. My problem is, that this metric seems to not have any value for when there are no failed connections (no zero, just emptiness) as seen below: This means, if there are failed connections, the metric recognizes them and my alert is fired, but the alert will never actually moves to "Resolved" state, because somehow as soon as the failed connections reach zero, there is no value to compare the threshold with and the alert stays open indefinitely, even though there are no failed connections on the database. Is this common knowledge and should be like this by design? And if yes, how can I actually modify my alert to be "Resolved" once there is no value in the metric table? I thank anyone who can help me with this issue in advance.

The Azure Database for PostgreSQL -flexible server metric connections_failed doesn't always have a value

Zeinab Khosravi 40

Hello everyone,

I have faced an issue regarding Azure Database for PostgreSQL -flexible server metrics, which I hoped someone here can help me solve it.

I have an Azure Database for PostgreSQL - flexible server, which has an alert configured on it, so I will be notified if the number of failed connections on the server exceeds a threshold, it uses the connections_failed metric, that is supposedly enabled by default. My problem is, that this metric seems to not have any value for when there are no failed connections (no zero, just emptiness) as seen below:

Screenshot 2024-03-11 at 12.42.51

This means, if there are failed connections, the metric recognizes them and my alert is fired, but the alert will never actually moves to "Resolved" state, because somehow as soon as the failed connections reach zero, there is no value to compare the threshold with and the alert stays open indefinitely, even though there are no failed connections on the database.

Is this common knowledge and should be like this by design? And if yes, how can I actually modify my alert to be "Resolved" once there is no value in the metric table?

I thank anyone who can help me with this issue in advance.

GeethaThatipatri-MSFT 27,887 Reputation points Microsoft Employee

2024-03-11T20:40:14.45+00:00

Hi, @Shadi Thanks for posting your question in the Microsoft Q&A forum.

it is common for metrics to not have a value when there are no failed connections. This is because the metric only records data when there is an event to record.

Here is the document on how to create and manage alerts in Azure Monitor, including how to create alert rules based on metric values, how to configure alert thresholds, and how to manage alert notifications.

Create new alert rule that triggers when the connections_failed metric is equal to zero https://learn.microsoft.com/en-us/azure/azure-monitor/alerts/alerts-create-metric-alert-rule#create-an-alert-rule-based-on-metric-values

Regards

Geetha
Zeinab Khosravi 40 Reputation points

2024-03-12T08:57:44.01+00:00

Hello @GeethaThatipatri-MSFT

thank you for your response. I might be wrong, but unfortunately I don't think that that really helps me!

First of all, my threshold is actually set to greater than 30 for the alert, because I know sometimes the developers make a mistake in their password or something and their request connection fails and I don't want the alert to be so noisy in those cases. Also seeing as how the metric doesn't recognize "zero" as a value, I don't know how setting the threshold to zero might help.

Also, I have checked some other metrics on postgresql flexible server, that are either enabled by default like connections_failed metric or have been enabled using server parameters, and have zero as their value and in their case, this zero is shown in the graph and inside AzureMetrics table. For example read_iops, disk_queue_depth, temp_files, deadlocks, and so on. So the metric graph for them looks like this:

and using this "zero" value, a similar alert wouldn't have the same issue as the connections_failed metric has!

The only metrics with similar behavior are the PgMetrics, but that is to be expected, since they are disabled by default and we haven't enabled them for our setup!

Any help here would be highly appreciated.
Zeinab Khosravi 40 Reputation points

2024-04-08T12:22:47.79+00:00

Hello @GeethaThatipatri-MSFT

I was wondering if you or any other of your colleagues have any more information/suggestion this topic.

Thank you in advance.
GeethaThatipatri-MSFT 27,887 Reputation points Microsoft Employee

2024-04-11T23:51:19.0066667+00:00

@Zeinab Khosravi Apologies for not responding on this

Default enabled = Yes for Failed Connections please check the documentation Monitoring and metrics - Azure Database for PostgreSQL - Flexible Server | Microsoft Learn

alert will display based on granularity selected time period, Please see the configuration if you can find any difference check if the Greater than or equal to 1 have same behavior?

Regards

Geetha

Share via

The Azure Database for PostgreSQL -flexible server metric connections_failed doesn't always have a value