Github user squito commented on a diff in the pull request:

    https://github.com/apache/spark/pull/8760#discussion_r43798070
  
    --- Diff: docs/configuration.md ---
    @@ -1141,6 +1141,48 @@ Apart from these, the following properties are also 
available, and may be useful
       </td>
     </tr>
     <tr>
    +  <td><code>spark.scheduler.blacklist.enabled</code></td>
    +  <td>true</td>
    +  <td>
    +    If set to ture, executor blacklist feature will be enabled to avoid 
allocte new task on bad executor. The logic to define bad executor is based on 
BlacklistStrategy which is also configurable.
    +  </td>
    +</tr>
    +<tr>
    +  <td><code>spark.scheduler.executorTaskBlacklistTime</code></td>
    +  <td>0L</td>
    +  <td>
    +    The threshold to deside blacklist executor, if last failed time is 
older than current time subtract executorTaskBlacklistTime, then the executor 
could be removed from blacklist.
    +  </td>
    +</tr>
    +<tr>
    +  <td><code>spark.scheduler.blacklist.recoverPeriod</code></td>
    +  <td>60L</td>
    +  <td>
    +    The period between runnning blacklist recover process. 
    +  </td>
    +</tr>
    +<tr>
    +  <td><code>spark.scheduler.blacklist.strategy</code></td>
    +  <td>default</td>
    +  <td>
    +    The strategy to determine executor blacklist and node blacklist. There 
are "default" strategy which is the srandard behavior before spark 1.6 and 
"threshold" strategy which is taskId unrelated and can learn experience from 
other taskSet to avoid allocating tasks on problematic executors.
    --- End diff --
    
    this definitely needs more thorough documentation ... probably more than 
should go into the description of a parameter. Maybe it deserves an additional 
section like 
http://spark.apache.org/docs/latest/job-scheduling.html#dynamic-resource-allocation
 ?
    I will think about this a bit more ...


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to