accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sean Busbey" <s...@manvsbeard.com>
Subject Re: Review Request 18444: ACCUMULO-2399 Continuous* wait a bit for scanners
Date Tue, 25 Feb 2014 01:20:15 GMT


> On Feb. 25, 2014, 12:38 a.m., Josh Elser wrote:
> > src/server/src/main/java/org/apache/accumulo/server/test/continuous/ContinuousUtil.java,
line 46
> > <https://reviews.apache.org/r/18444/diff/1/?file=502880#file502880line46>
> >
> >     Might as well just sleep/check/repeat indefinitely. The stop script just pkill's
the process.
> 
> Sean Busbey wrote:
>     That will further hide the error though. Especially for tests that run for 3 days,
I want failures to happen up front and loudly. Because, by necessity, most people are going
fire and then check back on logs and such after the elapsed time is up. Schedules get real
hard when you sink a whole 3 day turn around.
> 
> Josh Elser wrote:
>     I would've assumed that you check to make sure things are running before you ignore
it for a long period of time, but that's just me :P. If we're going to be touching this code,
just thinking about making something more robust than a "wait 3 seconds" because I know my
attention span and it can certainly exceed that which makes it no better than not waiting.

Depends on how long you wait though. What if the table creation is blocked waiting on some
FATE operation for 30 seconds? 5 minutes? It's just better to make things have easily observable
state. Creating the table ahead of time in the shell makes it very obvious if things are correct
(or if you are blocked waiting).


- Sean


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/18444/#review35356
-----------------------------------------------------------


On Feb. 24, 2014, 11:35 p.m., Mike Drob wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/18444/
> -----------------------------------------------------------
> 
> (Updated Feb. 24, 2014, 11:35 p.m.)
> 
> 
> Review request for accumulo.
> 
> 
> Bugs: ACCUMULO-2399
>     https://issues.apache.org/jira/browse/ACCUMULO-2399
> 
> 
> Repository: accumulo
> 
> 
> Description
> -------
> 
> ACCUMULO-2399 Continuous* wait a bit for scanners
> 
> When starting ContinuousIngest concurrently with consumers, many of the
> consumers would die if they came up before the table had been created.
> Make the code a bit more robust during start-up.
> 
> 
> Diffs
> -----
> 
>   src/server/src/main/java/org/apache/accumulo/server/test/continuous/ContinuousBatchWalker.java
4659affc49fc7fc2519e092330dda817ccbbadcd 
>   src/server/src/main/java/org/apache/accumulo/server/test/continuous/ContinuousQuery.java
c8ae6ecf9e5ef2bd51cb803cba429baaf57f3d6a 
>   src/server/src/main/java/org/apache/accumulo/server/test/continuous/ContinuousScanner.java
0ac3df63a260a201544b3623d4901b355d7ea21a 
>   src/server/src/main/java/org/apache/accumulo/server/test/continuous/ContinuousUtil.java
PRE-CREATION 
>   src/server/src/main/java/org/apache/accumulo/server/test/continuous/ContinuousWalk.java
7d1e7f95c7e5a7a172990322e2992e536f2725e3 
> 
> Diff: https://reviews.apache.org/r/18444/diff/
> 
> 
> Testing
> -------
> 
> Unit tests. Started processes on 3 node cluster and batch walkers did not die.
> 
> 
> Thanks,
> 
> Mike Drob
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message