Hi Pritam, Slots in TM focus more on logical management for cores, this means that TM could not assign physical cores to each slot. RM will manage the core usage of each TM, such as TM1 using 2 cores and TM2 using 3 cores. According to the resource usage required by the task (also logical and estimated), RM allocates slots from TM for the task. During the execution phase, whether the slot really uses so many cores, or not, or exceeds the allocated quantity, TM will not control it.
Currently, `Fine-Grained Resource Managemeng` should be the default resource management on the main branch of Flink, you can refer to doc[1] for more information. Returning to your question, if you find a bottleneck in the TM CPU usage of the job, you can try increasing the number of TM cores, which will improve job performance. [1] https://nightlies.apache.org/flink/flink-docs-master/docs/deployment/finegrained_resource/ Best, Shammon FY On Wed, Jun 21, 2023 at 3:18 AM Pritam Agarwala <pritamagarwala...@gmail.com> wrote: > Hi Team, > > I have a confusion about how TM's core will be used for task slots. > > I have TMs with 3 cores and 4 task slots . Each task slot will be using > one core to perform a task and one task slot has to wait to get resources ? > If I increase the cores to 4 , same as task slots per TM will it help to > improve performance ?? Is there any thumb rule ? > > > Thanks & Regards, > Pritam >
