statsd | An efficient Statsd Go client | Analytics library

I believe that your issue is that you appear to be issuing the PRAGMA only in the onCreate callback.

This callback will only be called the once and I don't believe that the state persists. As such RECURSIVE_TRIGGERS is ON (left ON).

I'd suggest adding an onOpen callback to execute the PRAGMA.

In order to install airflow in repeatable way you need to follow the approach with constraints: http://airflow.apache.org/docs/apache-airflow/stable/installation.html#installation-script - note that you are using rather old version of Airflow so 1.10.9 constraints will be rather old, I'd recommend you to upgrade to later version of Airflow.

You can also prepare such a constraints file yourself from your installation pip freeze > constraints.txt and then you can modify the file and set the pandas_gbq to 0.14.1 (and then use that constraint file with --constraint flag). This will give you the exact versions of the dependencies you already have and force installation of 0.14.1 for pandas-gbq

I think I found a root cause of this issue. In airflow helm chart I see this code:

We figured this out in the comments, I'm posting an answer that summarizes it all up: it seems the root cause was using different versions of different spring-boot modules.

It is a good rule of thumb to not define the versions yourself but use BOMs and let them define the versions for you, e.g. see: spring-boot-dependencies. This way you will use the compatible (and tested) versions.

management.metrics.tags.your-tag is the way to add tags to all of your metrics. A good way to check this is looking at /actuator/metrics.

Here's another take on the IngressRoute, adapted to your environment.

I think 99% of the issue is actual route matching, especially if you say --api.insecure works. Also as a rule of a thumb, logging & access log would help a lot in the DaemonSet definition.

Replace

Micrometer uses MeterFilters registered with a MeterRegistry to modified the meters that are registered. The modifications include the ability to map a meter's ID to something different.

In Spring Boot, you can use a MeterRegistryCustomizer bean to add a MeterFilter to a registry. You can use generics to work with a registry of a specific type, for example MeterRegistryCustomizer for a customizer that is only interested in customizing the Datadog registry.

Putting this together, you can map the ID of the http.server.request meter to i.want.to.be.different using the following bean:

This answer is the synthesis of the discussion we had in the comment section which eventually led to a solution.

Before we do anything in such a situation, it is vital to double-check whether there was some change after all that we did not see for some reason. It is important not to omit this double-check, because this does not take much time, while the systematic approach that leads to the solution involves a lot of effort. So, with this double-check we have a lot of potential gain with virtually no risk/cost. I personally use git as a version control system, but any system that is reliably showing such differences will do the job.

Seeing many not working image buttons looks overwhelming, we need to focus our research to a manageable and understandable problem-space. So, we need to choose one of the not working image buttons and find out why that does not work. If we are lucky, then solving that problem will either automatically solve the others, or give us some strong hints for the others. Naturally, we could have multiple very different problems, which off course means that we need to repeat the process outlined in this answer.

A very important hint is that not all the image buttons show this behavior, so we know that the problem is unlikely to be related to image buttons and it's more likely related to actions triggered by events on the image buttons.

Some debugging and logging is always helpful. Debugging shows us what happens, so if we do debug, we might find some bad behavior at some point which might well be the issue itself. If we do some logging, then we can see what happens while the program is running. We can do either or both.

In some cases there is no solution. For example a remotely working API might stop working, which is an unsolvable issue. In this case detecting this situation is vital. If there is no solution, then at least we need to find this out as quickly as possible, so we do not sweat on it too much.

Some steps towards the solution:

1. Reduce the problem space to a small, manageable size (and possibly provide more information into your question)
2. Define your problem well (Currently you link your problem to image buttons, but you are wrong to do so, since the fact that some image buttons work disproves the hypothesis that image buttons would work wrongly)
3. Look into the undivideable units of work, in your case the actions, choose one of them as a current focus and redefine the problem into an even smaller size)
4. Log and debug to collect information
5. From the list of the information you have collected plan a few experiments
6. Sequentially try all your hypotheses
7. If the solution is not yet found, then you might start all over from point 1. or you might decide to remove the action (temporarily) and gradually implement it back, until you either find what the problem is or accidentally solved it

With help of @myeongkil kim I was able to solve this issue. The solution was I just had to remove all the docker volumes by docker volume prune and after this, bringing the network back up solves this problem.