23.07.2014 Views

Lustre 1.6 Operations Manual

Lustre 1.6 Operations Manual

Lustre 1.6 Operations Manual

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

pid inode goal result foundgrps cr \ merge tailbroken<br />

2838 32801 4/12288/1 5/925/1 31 2 1 \ 0 0<br />

2838 32802 4/12288/1 5/926/1 31 2 1 \ 1 2<br />

2838 32803 4/12288/1 5/927/1 31 2 1 \ 0 0<br />

2838 32804 4/12288/1 5/928/1 31 2 1 \ 1 32<br />

2838 32805 4/12288/1 5/929/1 31 2 1 \ 0 0<br />

2838 32806 4/12288/1 5/930/1 31 2 1 \ 1 2<br />

2838 32807 4/12288/1 5/931/1 31 2 1 \ 0 0<br />

2838 24579 3/12288/1 3/12289/1 11 1 1 \ 0 0<br />

The parameters are described below:<br />

Parameter<br />

pid<br />

inode<br />

goal<br />

result<br />

found<br />

grps<br />

cr<br />

queue<br />

merge<br />

tail<br />

broken<br />

Description<br />

Process that made the allocation.<br />

inode number allocated blocks<br />

Initial request that came to mballoc (group/block-in-group/number-of-blocks)<br />

What mballoc actually found for this request.<br />

Number of free chunks mballoc found and measured before the final decision.<br />

Number of groups mballoc scanned to satisfy the request.<br />

Stage at which mballoc found the result:<br />

0 - best in terms of resource allocation. The request was 1MB or larger and was<br />

satisfied directly via the kernel buddy allocator.<br />

1 - regular stage (good at resource consumption)<br />

2 - fs is quite fragmented (not that bad at resource consumption)<br />

3 - fs is very fragmented (worst at resource consumption)<br />

Total bytes in active/queued sends.<br />

Whether the request hit the goal. This is good as extents code can now merge<br />

new blocks to existing extent, eliminating the need for extents tree growth.<br />

Number of blocks left free after the allocation breaks large free chunks.<br />

How large the broken chunk was.<br />

Most customers are probably interested in found/cr. If cr is 0 1 and found is less<br />

than 100, then mballoc is doing quite well.<br />

Also, number-of-blocks-in-request (third number in the goal triple) can tell the<br />

number of blocks requested by the obdfilter. If the obdfilter is doing a lot of small<br />

requests (just few blocks), then either the client is processing input/output to a lot of<br />

small files, or something may be wrong with the client (because it is better if client<br />

sends large input/output requests). This can be investigated with the OSC<br />

rpc_stats or OST brw_stats mentioned above.<br />

Chapter 20 <strong>Lustre</strong>Proc 20-21

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!