<div class="gmail_quote">[Sorry I'm resending this message because I didn't follow the title convention]</div><div class="gmail_quote"><br>Hello,<div><br></div><div>I'm a newbie at the GlusterFS. </div><div>>From the distribute translator document, <a href="http://www.gluster.org/docs/index.php/Understanding_DHT_Translator" target="_blank">http://www.gluster.org/docs/index.php/Understanding_DHT_Translator</a></div>
<div><br></div><div>It seems to calculate hash value based on its parent directory.</div><div><br></div><div>For example, I have 3 bricks, br1, br2, br3 and each has 100G.</div><div>And I'm going to more files on the following two directories, one is 10G and the other is 150G</div>
<div><br></div><div>/aa/* (thousand files total 10G)</div><div>/bb/* (tens of thousand files total 150G)</div><div><br></div><div>If files in /aa/ is mapped to br1 and /bb/ is mapped to br2, would the overflowing 50G of /bb/ be stored at one of free spaces in br1 or br3?</div>
<div>(min-free-disk is 0 for easier example)</div><div><br></div><div>If it is supported, how GFS can fine the overflowing files with O(1).</div><div>Otherwise, total size of a directory might be much smaller than one brick size. (many directories can use same brick)</div>
<div><br></div><font color="#888888"><div><a href="http://www.gluster.org/docs/index.php/Understanding_DHT_Translator" target="_blank"></a><div>Min</div></div>
</font></div><br><div>Thanks</div>