On 25 Apr 2019, at 17:29, Bogdan-Andrei Iancu wrote: > Hi Dan, > > The load is considered as the time a processes is busy in doing something > (instead of waiting for a new task to handle). It is completely unrelated to > the CPU load. Also the load will count any busy waiting or I/O waiting done > by the process.
In this case is it correct to define load per process as time_spent_processing_per_cycle/measurement_cycle? Also regarding this measurement cycle, I see a loose reference in the docs mentioning that a cycle is "like 2 seconds". Can the cycle length be defined, and if so how? If it cannot be defined, what is its actual value? (the "like 2 seconds" reference sounds vague). > > Regards, > > Bogdan-Andrei Iancu > > OpenSIPS Founder and Developer > https://www.opensips-solutions.com > OpenSIPS Summit 2019 > https://www.opensips.org/events/Summit-2019Amsterdam/ > > On 04/25/2019 04:47 PM, Dan Pascu wrote: >> I'm trying to understand how autoscaling works and I'd like some >> clarifications. >> >> When a process group load is calculated is that based on actual CPU load, or >> is it just computed as busy_processes/total_processes in that group? >> >> What I'm trying to figure out is the behavior in the case of the TCP process >> group, which is let's say configured to spawn a new worker when it reaches >> 70% load for the TCP worker group. Now consider that this group starts with >> 5 workers and at some point in time, all 5 workers are busy processing a >> message. In addition let's consider that the DNS lookup is misconfigured and >> it takes 3 seconds to get an answer. This means that if the proxy receives 5 >> messages at a time, all 5 workers will be blocked in a synchronous DNS >> lookup for 3 seconds, but each using 0% CPU. >> >> My question is, in this 3 second time window (when all 5 workers are blocked >> in waiting for the DNS lookup), if a new request arrives over TCP, will >> opensips spawn a new TCP worker because all 5 workers are busy and it >> considers the load to be 100%, or it will do nothing because it uses CPU >> load which is 0% since all 5 processes are sleeping waiting for the DNS >> answer? >> >> -- >> Dan >> >> >> >> >> >> _______________________________________________ >> Users mailing list >> [email protected] >> http://lists.opensips.org/cgi-bin/mailman/listinfo/users > -- Dan _______________________________________________ Users mailing list [email protected] http://lists.opensips.org/cgi-bin/mailman/listinfo/users
