github.com/iqoqo/nomad@v0.11.3-0.20200911112621-d7021c74d101/website/pages/docs/schedulers.mdx (about)

     1  ---
     2  layout: docs
     3  page_title: Schedulers
     4  sidebar_title: Schedulers
     5  description: Learn about Nomad's various schedulers.
     6  ---
     7  
     8  # Schedulers
     9  
    10  Nomad has three scheduler types that can be used when creating your job:
    11  `service`, `batch` and `system`. Here we will describe the differences between
    12  each of these schedulers.
    13  
    14  ## Service
    15  
    16  The `service` scheduler is designed for scheduling long lived services that
    17  should never go down. As such, the `service` scheduler ranks a large portion
    18  of the nodes that meet the job's constraints and selects the optimal node to
    19  place a task group on. The `service` scheduler uses a best fit scoring algorithm
    20  influenced by Google's work on [Borg]. Ranking this larger set of candidate
    21  nodes increases scheduling time but provides greater guarantees about the
    22  optimality of a job placement, which given the service workload is highly
    23  desirable.
    24  
    25  Service jobs are intended to run until explicitly stopped by an operator. If a
    26  service task exits it is considered a failure and handled according to the job's
    27  [restart] and [reschedule] stanzas.
    28  
    29  ## Batch
    30  
    31  Batch jobs are much less sensitive to short term performance fluctuations and
    32  are short lived, finishing in a few minutes to a few days. Although the `batch`
    33  scheduler is very similar to the `service` scheduler, it makes certain
    34  optimizations for the batch workload. The main distinction is that after finding
    35  the set of nodes that meet the job's constraints it uses the power of two
    36  choices described in Berkeley's [Sparrow] scheduler to limit the number of nodes
    37  that are ranked.
    38  
    39  Batch jobs are intended to run until they exit successfully. Batch tasks that
    40  exit with an error are handled according to the job's [restart] and [reschedule]
    41  stanzas.
    42  
    43  ## System
    44  
    45  The `system` scheduler is used to register jobs that should be run on all
    46  clients that meet the job's constraints. The `system` scheduler is also invoked
    47  when clients join the cluster or transition into the ready state. This means
    48  that all registered `system` jobs will be re-evaluated and their tasks will be
    49  placed on the newly available nodes if the constraints are met.
    50  
    51  This scheduler type is extremely useful for deploying and managing tasks that
    52  should be present on every node in the cluster. Since these tasks are
    53  managed by Nomad, they can take advantage of job updating,
    54  service discovery, and more.
    55  
    56  Since Nomad 0.9, the system scheduler will preempt eligible lower priority
    57  tasks running on a node if there isn't enough capacity to place a system job.
    58  See [preemption] for details on how tasks that get preempted are chosen.
    59  
    60  Systems jobs are intended to run until explicitly stopped either by an operator
    61  or [preemption]. If a system task exits it is considered a failure and handled
    62  according to the job's [restart] stanza; system jobs do not have rescheduling.
    63  
    64  [borg]: https://research.google.com/pubs/pub43438.html
    65  [sparrow]: https://cs.stanford.edu/~matei/papers/2013/sosp_sparrow.pdf
    66  [preemption]: /docs/internals/scheduling/preemption
    67  [restart]: /docs/job-specification/restart
    68  [reschedule]: /docs/job-specification/reschedule