about cpu cores
given my spark cluster has 128 cores totally. If the jobs (each job was assigned only one core) I submitted to the cluster are over 128, what will happen? Thank you.
Re: about cpu cores
Jobs consist of tasks, each of which consumes a core (can be set to >1 too, but that's a different story). If there are more tasks ready to execute than available cores, some tasks simply wait. On Sun, Jul 10, 2022 at 3:31 AM Yong Walt wrote: > given my spark cluster has 128 cores totally. > If the jobs (each job was assigned only one core) I submitted to the > cluster are over 128, what will happen? > > Thank you. >
Re: about cpu cores
Mainly depends what your cluster manager Yarn or kubernates ? Best Tufan On Sun, 10 Jul 2022 at 14:38, Sean Owen wrote: > Jobs consist of tasks, each of which consumes a core (can be set to >1 > too, but that's a different story). If there are more tasks ready to > execute than available cores, some tasks simply wait. > > On Sun, Jul 10, 2022 at 3:31 AM Yong Walt wrote: > >> given my spark cluster has 128 cores totally. >> If the jobs (each job was assigned only one core) I submitted to the >> cluster are over 128, what will happen? >> >> Thank you. >> >
Re: about cpu cores
We were using Yarn. thanks. On Sun, Jul 10, 2022 at 9:02 PM Tufan Rakshit wrote: > Mainly depends what your cluster manager Yarn or kubernates ? > Best > Tufan > > On Sun, 10 Jul 2022 at 14:38, Sean Owen wrote: > >> Jobs consist of tasks, each of which consumes a core (can be set to >1 >> too, but that's a different story). If there are more tasks ready to >> execute than available cores, some tasks simply wait. >> >> On Sun, Jul 10, 2022 at 3:31 AM Yong Walt wrote: >> >>> given my spark cluster has 128 cores totally. >>> If the jobs (each job was assigned only one core) I submitted to the >>> cluster are over 128, what will happen? >>> >>> Thank you. >>> >>
Re: about cpu cores
so as an average every 4 core , you get back 3.6 core in Yarn , but you can use only 3 . in Kubernetes you get back 3.6 and also can use 3.6 Best Tufan On Mon, 11 Jul 2022 at 11:02, Yong Walt wrote: > We were using Yarn. thanks. > > On Sun, Jul 10, 2022 at 9:02 PM Tufan Rakshit wrote: > >> Mainly depends what your cluster manager Yarn or kubernates ? >> Best >> Tufan >> >> On Sun, 10 Jul 2022 at 14:38, Sean Owen wrote: >> >>> Jobs consist of tasks, each of which consumes a core (can be set to >1 >>> too, but that's a different story). If there are more tasks ready to >>> execute than available cores, some tasks simply wait. >>> >>> On Sun, Jul 10, 2022 at 3:31 AM Yong Walt wrote: >>> given my spark cluster has 128 cores totally. If the jobs (each job was assigned only one core) I submitted to the cluster are over 128, what will happen? Thank you. >>>
Re: about cpu cores
Hi, please see Sean's answer and please read about parallelism in spark. Regards, Gourav Sengupta On Mon, Jul 11, 2022 at 10:12 AM Tufan Rakshit wrote: > so as an average every 4 core , you get back 3.6 core in Yarn , but you > can use only 3 . > in Kubernetes you get back 3.6 and also can use 3.6 > > Best > Tufan > > On Mon, 11 Jul 2022 at 11:02, Yong Walt wrote: > >> We were using Yarn. thanks. >> >> On Sun, Jul 10, 2022 at 9:02 PM Tufan Rakshit wrote: >> >>> Mainly depends what your cluster manager Yarn or kubernates ? >>> Best >>> Tufan >>> >>> On Sun, 10 Jul 2022 at 14:38, Sean Owen wrote: >>> Jobs consist of tasks, each of which consumes a core (can be set to >1 too, but that's a different story). If there are more tasks ready to execute than available cores, some tasks simply wait. On Sun, Jul 10, 2022 at 3:31 AM Yong Walt wrote: > given my spark cluster has 128 cores totally. > If the jobs (each job was assigned only one core) I submitted to the > cluster are over 128, what will happen? > > Thank you. >