I was testing ocfs2 on a 2 node cluster set up.<br><br>ocfs2-tools version is 1.6.3<br>ocfs2 kernel version is 2.6.36<br><br>Using cman on 2 nodes<br><br>node02 dw # ping_pong -rwm /data/test.dat 3<br>data increment = 2<br>
14 locks/sec<br><br>node01 dw # ping_pong -rw /data/test.dat 3<br>data increment = 2<br> 10 locks/sec<br><br>node02 dw # ping_pong -r /data/test.dat 3<br> 1980 locks/sec<br><br>Using cman on 1 node<br><br>node02 dw # ping_pong -rwm /data/test.dat 3<br>
data increment = 1<br> 3165 locks/sec<br><br>node02 dw # ping_pong -rw /data/test.dat 3<br>data increment = 1<br> 3289 locks/sec<br><br>node02 dw # ping_pong -r /data/test.dat 3<br> 3267 locks/sec<br><br>The above seems reasonable and behaved as expected when the program was stopped and started repeatedly on either node. So it passed the test.<br>
<br><br><br>Unfortunately o2cb didnt behave as expected.<br><br>Using o2cb on 2 nodes<br><br>node02 dw # ping_pong -rwm /data/test.dat 3<br>data increment = 199<br>data increment = 76<br>data increment = 1<br>data increment = 117<br>
data increment = 84<br>data increment = 1<br>data increment = 212<br>data increment = 211<br>data increment = 1<br>data increment = 225<br>data increment = 85<br>data increment = 1<br>data increment = 237<br>data increment = 158<br>
data increment = 1<br>data increment = 75<br>data increment = 235<br>data increment = 1<br>data increment = 199<br>data increment = 4<br>data increment = 1<br> 171803 locks/sec<br><br>and occasionally you will get the data increment remaining at 1 when 2 nodes have joined<br>
<br>node02 dw # ping_pong -rw /data/test.dat 3<br>data increment = 2<br> 45 locks/sec<br><br>node01 dw # ping_pong -r /data/test.dat 3<br> 2155175 locks/sec<br><br>Using cman on 1 node<br><br>node02 dw # ping_pong -rwm /data/test.dat 3<br>
data increment = 1<br> 2230553 locks/sec<br><br>node02 dw # ping_pong -rw /data/test.dat 3<br>data increment = 1<br> 299492 locks/sec<br><br>node02 dw # ping_pong -r /data/test.dat 3<br> 2249787 locks/sec<br><br>The mmap test failed. The read/write test 'usually' worked but frequently didn't increment the 'data increment' to 2, when the program was running on both nodes.<br>
<br>Any suggestions why this isn't working as expected?<br><br>Thanks, Dan<br><br>