[Gluster-devel] glusterfs stripe

[Raghavendra G]

Hi john

On Thu, Jul 17, 2008 at 5:09 PM, John Marshall <John.Marshall at ec.gc.ca>
wrote:

> Raghavendra G wrote:
>
>> Hi John,
>>
>> Please find the inlined comments.
>>
>> On Thu, Jul 17, 2008 at 2:53 AM, John Marshall <John.Marshall at ec.gc.ca<mailto:
>> John.Marshall at ec.gc.ca>> wrote:
>>
>>    Hi,
>>
>>    I was testing out the glusterfs "type cluster/stripe" and saw in
>>    the output:
>>    -----
>>    2008-07-16 22:48:12 D [fuse-bridge.c:1604:fuse_writev_cbk]
>>    glusterfs-fuse: 169275: WRITE => 131072/131072,21474185188/21474316260
>>    2008-07-16 22:48:12 D [fuse-bridge.c:1641:fuse_write]
>>    glusterfs-fuse: 169276: WRITE (0x80560b0, size=131072,
>>    offset=21474316260)
>>    -----
>>    Whether the client is configured with "option block-size *:4MB" or
>>    "option block-size *:1MB" does not seem to change what appears
>>    above to be write() operations of 128KB at a time. Am I understanding
>>    this correctly?
>>
>>
>> "option block-size " is specific to cluster/stripe xlator. It determines
>> the size of the chunks the file has to be sliced into and distributed among
>> stripe's children. It does not affect the sizes in which fuse does
>> writes/reads.
>>
> Is there any way to change the size of the writes/reads? Have you done any
> tests to see the impact of going to a larger value than 128KB?


glusterfs is using the maximum  block-size allowed by fuse. Also this
block-size is not configurable in glusterfs.


>
> As a side question, must the block-size remain the same for the life of the
> filesystem? I.e., if the fs was used with block-size of 4MB, then mounted
> later with 1MB, would this mess things up?
>
> Thanks,
> John
>



-- 
Raghavendra G

A centipede was happy quite, until a toad in fun,
Said, "Prey, which leg comes after which?",
This raised his doubts to such a pitch,
He fell flat into the ditch,
Not knowing how to run.
-Anonymous