Closed yannliepallotools closed 3 months ago
Hi @yannliepallotools, thanks for your question! I was just wondering if you could provide more information about how the talk_interval
is affecting your system load. So far we have not seen agent HTTP requests have a significant performance impact.
I just want to gather as much information about the underlying issue.
It's not really a load problem but because we are doing some big computation on our server we wanted to buffer the access log file to limit the disk IOs. If we do it some datas weren't send, ie. the request time. But maybe it has nothing to do with the talk_interval.
Yeah, disk IOPS by the agent won't be affected by the talk interval. We don't expose the collection intervals at the moment, but if agent IOPS are an issue you may consider configuring the agent to use syslog
instead (https://amplify.nginx.com/docs/guide-installing-and-managing-nginx-amplify-agent.html#configuring-syslog). You can also configure agent to ignore access log files (https://amplify.nginx.com/docs/guide-installing-and-managing-nginx-amplify-agent.html#excluding-certain-nginx-log-files).
Would that work for you?
The problem is that if we disable the access log metrics we won't have access to the request time log. We'll let it be like this for the moment and monitor the disk IOs, if it's too heavy we'll disable access logging.
@yannliepallotools all metrics Amplify collects from the access log can be collected via syslog
. NGINX supports logging for access and error to syslog
: http://nginx.org/en/docs/syslog.html . If you want to keep the log files but just move Amplify collection to syslog
, you can configure NGINX to log to both a file as well as to syslog
and configure the Amplify agent to ignore the files.
Is it possible to change the talk_interval to reduce the load ?