I'm looking for some help to understand why write throughput in my Lustre IB cluster is only about 50-80 MB/sec, while read performance is 5-6 GB/sec.  I'm running 2.4.1-RC2-PRISTINE on my MDT and 2 OSS's.  Also using 2.4.92 on my 16 clients, 3 of which run iozone write throughput tests.

 

I noticed a few threads in the 2.4.1 RC1 and RC2 timeframe discussing low write performance.  I noticed that curr_dirty_bytes start off at 0 at the start of the test as one would expect.  As the test proceeds, one OSS's curr_dirty_bytes stays pegged at some huge number, implying it didn't see a commit.  The other OSS's curr_dirty_bytes varies during the test as iozone writes data that gets committed.  What can I look at to see why the commit isn't happening?

 

Thanks in advance,

Michael