You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When dealing with multiple slaves in varying CPU/memory configurations, it would be great to enable relative scheduling - say, one server has 8 CPUs (150% load), the other has 2 CPUs (120% load) - so least_cpu should consider the 800% vs 200% CPU ceiling and schedule the next job for server 1, even though the absolute CPU load is higher. It would also be great to be able to define CPU / mem thresholds, where no new jobs are assigned (server 2 has less than 90% CPU available, so will receive no new jobs until dropping below 110% load) and expected cpu/mem load per job so jobs would remain in the queue if there are no sufficient ressources available. This would require to get the system CPU/mem load, not only the child process.
This would help us to distribute the load more evenly.
Cheers,
Bernhard
The text was updated successfully, but these errors were encountered:
Summary
When dealing with multiple slaves in varying CPU/memory configurations, it would be great to enable relative scheduling - say, one server has 8 CPUs (150% load), the other has 2 CPUs (120% load) - so least_cpu should consider the 800% vs 200% CPU ceiling and schedule the next job for server 1, even though the absolute CPU load is higher. It would also be great to be able to define CPU / mem thresholds, where no new jobs are assigned (server 2 has less than 90% CPU available, so will receive no new jobs until dropping below 110% load) and expected cpu/mem load per job so jobs would remain in the queue if there are no sufficient ressources available. This would require to get the system CPU/mem load, not only the child process.
This would help us to distribute the load more evenly.
Cheers,
Bernhard
The text was updated successfully, but these errors were encountered: