Re: [Kosmosfs-users] Missing Chunks with Hadoop

[Michael <mic...@gm...>]

Here is both MIN and MAX:

const float MIN_SERVER_SPACE_UTIL_THRESHOLD = 0.3;
const float MAX_SERVER_SPACE_UTIL_THRESHOLD = 0.9;

Thanks,
Michael

Sriram Rao wrote:
> Can you tell me the value you have for:
> 
> const float MAX_SERVER_SPACE_UTIL_THRESHOLD
> 
> in src/cc/meta/LayoutManager.cc
> 
> This controls how much space will be used on a given chunkserver node.
>  Even though you specify 12GB, we leave some space alone for other
> things on the node  and this value controls how much space to leave.
> 
> Sriram
> 
> On Mon, Dec 1, 2008 at 11:59 PM, Michael <mic...@gm...> wrote:
>> I figured it out!  It was hadoop or cptokfs (I realized it did in fact
>> have the same problem)  Even though my limit to the filesystem
>> chunkfolder was 12GB it would only upload about 7.9GB.  I increased this
>> to some ridiculous number (120GB) and it worked fine.  Does kfs overlook
>> this or take into consideration metadata/replications?
>>
>> Thanks,
>> Michael
>>
>> lohit wrote:
>>> Do you have output of bin/hadoop fs -put somefile newname?
>>> By any chance do you also have exit status of the command?
>>> Thanks,
>>> Lohit
>>>
>>> ------------------------------------------------------------------------
>>> *From:* Michael Miceli <mic...@gm...>
>>> *To:* Sriram Rao <sri...@gm...>
>>> *Cc:* kos...@li...
>>> *Sent:* Monday, December 1, 2008 6:20:01 PM
>>> *Subject:* Re: [Kosmosfs-users] Missing Chunks with Hadoop
>>>
>>> Yeah. It is no big deal for me because I can put the data onto the
>>> filesystem with cptokfs but
>>> bin/hadoop fs -put somefile newname
>>> Will not put all of the data on the kfs filesystem
>>>
>>>     On Dec 1, 2008 6:39 PM, "Sriram Rao" <sri...@gm...
>>>     <mailto:sri...@gm...>> wrote:
>>>
>>>     This is very odd. Can you tell me the sequence of commands you ran?
>>>
>>>     Sriram
>>>
>>>     On Mon, Dec 1, 2008 at 3:43 PM, Michael <mic...@gm...
>>>     <mailto:mic...@gm...>> wrote: > Hi again! > When usin...
>>>
>>>     >
>>>     -------------------------------------------------------------------------
>>>     > This SF.Net email is s...
>>>