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
Ceiling of "weight" (CPU/RAM) of processes on a worker.
BPMNs would be given a "weight" ahead of time, by a human. This could be determined by experimentation and testing of a workflow (means instead of maxes might want to be used here, but up to the human). If a weight is not available (not assigned), then a BPMN would just automatically be assigned a default weight. It would be just metadata about the BPMN. So somebody might say "edrgen.bpmn takes about 0.5 GB RAM, and uses about 30% CPU". So maybe there are two weights :
1). CPU worst case estimate,
2). memory worst case estimate.
And maybe you just can't exceed the machine's capacity by both of those measures.
The worker node would have to define a
maxCpu
maxMem.
When implementing this, one must be cognizant of the possibility that certain BPMNs may be starved out.
CURRENT PROJECT WORKAROUND: spend more money, and beef up worker instance classes, so that the worker can handle load.
The text was updated successfully, but these errors were encountered:
Ceiling of "weight" (CPU/RAM) of processes on a worker.
BPMNs would be given a "weight" ahead of time, by a human. This could be determined by experimentation and testing of a workflow (means instead of maxes might want to be used here, but up to the human). If a weight is not available (not assigned), then a BPMN would just automatically be assigned a default weight. It would be just metadata about the BPMN. So somebody might say "edrgen.bpmn takes about 0.5 GB RAM, and uses about 30% CPU". So maybe there are two weights :
1). CPU worst case estimate,
2). memory worst case estimate.
And maybe you just can't exceed the machine's capacity by both of those measures.
The worker node would have to define a
When implementing this, one must be cognizant of the possibility that certain BPMNs may be starved out.
CURRENT PROJECT WORKAROUND: spend more money, and beef up worker instance classes, so that the worker can handle load.
The text was updated successfully, but these errors were encountered: