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
Our scenario is as below: (1)T1 and T2 are two tasks (1)N1 and N2 are two nodes. I just want to know whether flow scheduler can get global optimized result if T1 has been scheduled on N1
--task_lib_dir is needed if you want to run tasks using Firmament's native cluster execution (in which we inject a dynamically linked library into the task binary). The parameter specifies the directory in which the compile .so file for this library resides (and hence it's in the build tree).
If you're using Firmament's native cluster execution, you can use the Python script in scripts/job/job_submit.py to submit a job with one or more tasks. You can also write your own script, as long as it sends the correct serialized protobufs.
Finally, it might be easiest to use Firmament via its Kubernetes integration in Poseidon. There, you can simply use well-established Kubernetes APIs to submit jobs.
I have some questions when running firmament? Could you help me?
docker run camsas/firmament:dev /firmament/build/src/coordinator --scheduler flow --flow_scheduling_cost_model 6 --listen_uri tcp:<host>:<port> --task_lib_dir=$(pwd)/build/src
Question 2: How to submit tasks and nodes info?
Our scenario is as below: (1)T1 and T2 are two tasks (1)N1 and N2 are two nodes. I just want to know whether flow scheduler can get global optimized result if T1 has been scheduled on N1
scenario
The text was updated successfully, but these errors were encountered: