Cron in http_poller still triggering after changing the pipeline configuration

[fbenaven (F B)]

Hi,

We have a pipeline configuration using some http_poller as inputs.

After changing the config from "cron" format (at '15 of all hours) to "every" (hourly basis), we see that the poller triggers hourly as intended but it also triggers at minute 15 as it was previously configured.

This happens in 2 machines, one running Ubuntu and other RHEL

Is the cron config cached somewhere? How can we get rid of it?

We tried both restarting the service and the full machine but the issue is still there.

In the logs we only see the new config ("every")

Thanks

[Badger]

Are you using config.reload_automatic, or are you restarting the logstash process. I am wondering if it is similar to this issue.

[fbenaven (F B)]

Thanks Badger for this.
We are required to restart the process to get the config changes, they are not loaded automatically.

The weird thing is that the issue remains both even if I reboot the machine and also if I remove the affected pipeline from the pipelines.yml file.

I don't know what to do to clean this up

[Badger]

Then I suspect that you are not running the configuration you think you are running. Can you try --log.level debug --config.debug?

[fbenaven (F B)]

It is reading it

in fact it is responding to changes we are doing in that file. The problem is it is doing things that were in previous versions of that file on top of the current ones. And they are related to the previous cron config.
It seems the cron config is cached somewhere and not deleted even it is not in the config file anymore

[fbenaven (F B)]

After reinstalling from scratch and seen that the symptoms were still there (and not believing in witches) I had to think in another possibilities. I finally discovered that a collegue of mine instantiated another logstash instance with the old config in a different environment pointing to the same elasticsearch cluster by mistake.
Thanks Badger for your support and sorry