luxolus/open-nomad
2
0
Fork 0
Code Issues 1 Pull requests Packages Releases Wiki Activity

Edits and links for the submitting a job guide

Browse source
This commit is contained in:
Michael Lange 2019-08-16 00:56:49 -07:00
parent d7261e20a8
commit 0b4c65b488
1 changed files with 11 additions and 12 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

23
website/source/guides/web-ui/submitting-a-job.html.md
View file

@ -19,7 +19,7 @@ the next step is to run the plan.
## Nomad plan ## Nomad plan
It is best practice to run `nomad plan` before running `nomad run`, so the Web UI enforcing this It is best practice to run `nomad plan` before running `nomad run`, so the Web UI enforces this
best practice. From the Job Run page, underneath the code editor, there is a Plan button. Clicking best practice. From the Job Run page, underneath the code editor, there is a Plan button. Clicking
this button will proceed the run process to the second step. this button will proceed the run process to the second step.
@ -37,8 +37,6 @@ potential issues early. Some potential issues are:
3. Your job has an unresolvable hard constraint (e.g., required port not available). 3. Your job has an unresolvable hard constraint (e.g., required port not available).
4. In order to start your job, other jobs must be preempted. 4. In order to start your job, other jobs must be preempted.
Read more about placement failures and preemptions below.
~> Screenshot (Job plan placement failures) ~> Screenshot (Job plan placement failures)
From the plan step, you can either cancel to make edits, or run the job. When you run the job, you From the plan step, you can either cancel to make edits, or run the job. When you run the job, you
@ -59,16 +57,16 @@ being no placement failures (e.g., artifact cannot download or container startup
## Preemptions ## Preemptions
Another class of potential issues when planning a job is preemptions. This happens when the cluster Another class of potential issues when planning a job is
does not have capacity for your job, but your job is a high priority and the cluster has preemptions [preemptions](/docs/internals/scheduling/preemption.html). This happens when the cluster does not
enabled. have capacity for your job, but your job is a high priority and the cluster has preemptions enabled.
~> Screenshot (Job plan preemptions) ~> Screenshot (Job plan preemptions)
Unlike with placement failures, when you submit a job that has expected preemptions, the job is will Unlike with placement failures, when you submit a job that has expected preemptions, the job will
start. However, other allocations will be stopped to free up capacity. start. However, other allocations will be stopped to free up capacity.
Note that with Nomad OSS, only system jobs can preempt allocations. Nomad Enterprise allows for both ~> With Nomad OSS, only system jobs can preempt allocations. Nomad Enterprise allows for both
service and batch type jobs to preempt lower priority allocations. service and batch type jobs to preempt lower priority allocations.
## Job Overview ## Job Overview
@ -78,11 +76,12 @@ Upon submitting a job, you will be redirected to the Job Overview page for the j
If this is a new job, the job will start in a queued state. If there are no placement failures, If this is a new job, the job will start in a queued state. If there are no placement failures,
allocations for the job will naturally transition from a starting to a running or failed state. allocations for the job will naturally transition from a starting to a running or failed state.
Nomad is quick to schedule allocations (i.e., find a client node to start the allocation on), but an Nomad is quick to schedule allocations (i.e., find a client node to start the allocation on), but an
allocation may sit in the starting state for awhile if it has to download source images or other allocation may sit in the starting state for awhile if it has to download
artifacts. It may also sit in a starting state if the task fails to start and requires retry [source images](/docs/job-specification/task.html#task-examples) or
attempts. [other artifacts](/docs/job-specification/artifact.html). It may also sit in a starting state if the
task fails to start and requires retry attempts.
If this is was an existing job that was resubmitted, the job overview will just show old allocations If this is was an existing job that was resubmitted, the job overview will show old allocations
moving into a completed status before new allocations are spun up. The exact sequence of events moving into a completed status before new allocations are spun up. The exact sequence of events
depends on the configuration of the job. depends on the configuration of the job.