The addition of versions of config files for various sizes of RAM in 1.4 is awesome.

That's all.

Aaron

Have you given them a test run? I was guessing them, so I'm not certain how
useful some of the smaller ones are.

Are there any other sizes you would like to see?

Sent from my phone, so pardon the typos and brevity.
On Mar 14, 2012 12:39 PM, "Aaron Cordova" <[EMAIL PROTECTED]> wrote:

> The addition of versions of config files for various sizes of RAM in 1.4
> is awesome.
>
> That's all.
>
> Aaron
>

Yes, as part of QA for the upcoming 1.4.0 release, each of the sizes was
tested for basic ingest/query/recovery.

And your guesses were not too bad: only the 512MB configuration failed. :-)

-Eric

On Wed, Mar 14, 2012 at 8:48 AM, John Vines <[EMAIL PROTECTED]> wrote:

> Have you given them a test run? I was guessing them, so I'm not certain
> how useful some of the smaller ones are.
>
> Are there any other sizes you would like to see?
>
> Sent from my phone, so pardon the typos and brevity.
> On Mar 14, 2012 12:39 PM, "Aaron Cordova" <[EMAIL PROTECTED]> wrote:
>
>> The addition of versions of config files for various sizes of RAM in 1.4
>> is awesome.
>>
>> That's all.
>>
>> Aaron
>>
>

2GB just failed: max tserver map memory 512MB and block cache sizes 115MB are too large for JVM config 256MB

On Mar 14, 2012, at 8:51 AM, Eric Newton wrote:

> Yes, as part of QA for the upcoming 1.4.0 release, each of the sizes was tested for basic ingest/query/recovery.
>
> And your guesses were not too bad: only the 512MB configuration failed. :-)
>
> -Eric
>
> On Wed, Mar 14, 2012 at 8:48 AM, John Vines <[EMAIL PROTECTED]> wrote:
> Have you given them a test run? I was guessing them, so I'm not certain how useful some of the smaller ones are.
>
> Are there any other sizes you would like to see?
>
> Sent from my phone, so pardon the typos and brevity.
>
> On Mar 14, 2012 12:39 PM, "Aaron Cordova" <[EMAIL PROTECTED]> wrote:
> The addition of versions of config files for various sizes of RAM in 1.4 is awesome.
>
> That's all.
>
> Aaron
>

Were you using the example for native code with compiling the native
libraries?

Sent from my phone, so pardon the typos and brevity.
On Mar 14, 2012 12:55 PM, "Aaron Cordova" <[EMAIL PROTECTED]> wrote:

> 2GB just failed: max tserver map memory 512MB and block cache sizes 115MB
> are too large for JVM config 256MB
>
>
>
> On Mar 14, 2012, at 8:51 AM, Eric Newton wrote:
>
> Yes, as part of QA for the upcoming 1.4.0 release, each of the sizes was
> tested for basic ingest/query/recovery.
>
> And your guesses were not too bad: only the 512MB configuration failed. :-)
>
> -Eric
>
> On Wed, Mar 14, 2012 at 8:48 AM, John Vines <[EMAIL PROTECTED]> wrote:
>
>> Have you given them a test run? I was guessing them, so I'm not certain
>> how useful some of the smaller ones are.
>>
>> Are there any other sizes you would like to see?
>>
>> Sent from my phone, so pardon the typos and brevity.
>> On Mar 14, 2012 12:39 PM, "Aaron Cordova" <[EMAIL PROTECTED]> wrote:
>>
>>> The addition of versions of config files for various sizes of RAM in 1.4
>>> is awesome.
>>>
>>> That's all.
>>>
>>> Aaron
>>>
>>
>
>

native, yes

There is enough memory, the config number just didn't add up. I upped the JVM max to 768MB in accumulo-env.sh and the tserver started. I don't know if that puts the max mem used by all procs over 2GB though. The map and cache sizes might need to be reduced for the 2GB config.

On Mar 14, 2012, at 8:58 AM, John Vines wrote:

> Were you using the example for native code with compiling the native libraries?
>
> Sent from my phone, so pardon the typos and brevity.
>
> On Mar 14, 2012 12:55 PM, "Aaron Cordova" <[EMAIL PROTECTED]> wrote:
> 2GB just failed: max tserver map memory 512MB and block cache sizes 115MB are too large for JVM config 256MB
>
>
>
> On Mar 14, 2012, at 8:51 AM, Eric Newton wrote:
>
>> Yes, as part of QA for the upcoming 1.4.0 release, each of the sizes was tested for basic ingest/query/recovery.
>>
>> And your guesses were not too bad: only the 512MB configuration failed. :-)
>>
>> -Eric
>>
>> On Wed, Mar 14, 2012 at 8:48 AM, John Vines <[EMAIL PROTECTED]> wrote:
>> Have you given them a test run? I was guessing them, so I'm not certain how useful some of the smaller ones are.
>>
>> Are there any other sizes you would like to see?
>>
>> Sent from my phone, so pardon the typos and brevity.
>>
>> On Mar 14, 2012 12:39 PM, "Aaron Cordova" <[EMAIL PROTECTED]> wrote:
>> The addition of versions of config files for various sizes of RAM in 1.4 is awesome.
>>
>> That's all.
>>
>> Aaron
>>
>

I'll test the 1GB on my laptop if the example files make it back into
the tar.gz file.

I've written a script to test all the configs.  I wasn't loading enough
data in test it properly before.

Looks like the 512MB and 1G configs need:

    <property>
      <name>logger.sort.buffer.size</name>
      <value>50M</value>
    </property>
-Eric

On Wed, Mar 14, 2012 at 9:53 AM, David Medinets <[EMAIL PROTECTED]>wrote:

> I'll test the 1GB on my laptop if the example files make it back into
> the tar.gz file.
>

Should tserver.walog.max.size also be adjusted down so that log sort
does not produce as many files?  The sort buffer size was cut by a
factor of 4, could adjust the max size by the same.

On Thu, Mar 15, 2012 at 9:20 AM, Eric Newton <[EMAIL PROTECTED]> wrote:
> I've written a script to test all the configs.  I wasn't loading enough data
> in test it properly before.
>
> Looks like the 512MB and 1G configs need:
>
>     <property>
>       <name>logger.sort.buffer.size</name>
>       <value>50M</value>
>     </property>
>
>
> -Eric
>
>
> On Wed, Mar 14, 2012 at 9:53 AM, David Medinets <[EMAIL PROTECTED]>
> wrote:
>>
>> I'll test the 1GB on my laptop if the example files make it back into
>> the tar.gz file.
>
>

My little test is ingesting 2M key/value pairs, which should be bigger than
memory on all 4 configurations.  After that, I restart all the processes
and verify the data.

A smaller log file size would make recoveries run faster.

-Eric

On Thu, Mar 15, 2012 at 9:30 AM, Keith Turner <[EMAIL PROTECTED]> wrote:

> Should tserver.walog.max.size also be adjusted down so that log sort
> does not produce as many files?  The sort buffer size was cut by a
> factor of 4, could adjust the max size by the same.
>
> On Thu, Mar 15, 2012 at 9:20 AM, Eric Newton <[EMAIL PROTECTED]>
> wrote:
> > I've written a script to test all the configs.  I wasn't loading enough
> data
> > in test it properly before.
> >
> > Looks like the 512MB and 1G configs need:
> >
> >     <property>
> >       <name>logger.sort.buffer.size</name>
> >       <value>50M</value>
> >     </property>
> >
> >
> > -Eric
> >
> >
> > On Wed, Mar 14, 2012 at 9:53 AM, David Medinets <
> [EMAIL PROTECTED]>
> > wrote:
> >>
> >> I'll test the 1GB on my laptop if the example files make it back into
> >> the tar.gz file.
> >
> >
>