3 Replies Latest reply on Jun 18, 2015 6:15 PM by Henk Vandenbergh-Oracle

    vdbench works for one IO rate but not the other

    2974002

      I have a situation, where vdbench works fine for 40K IOPs, but it fails if I change the same workload-file to 10K iops.

       

      Here is how the workload-file looks like:

       

      dedupratio=3

      dedupunit=4k

      compratio=4

      sd=default,openflags=directio,align=4k,threads=8

      sd=sd0,lun=/dev/dm-1

      sd=sd1,lun=/dev/dm-0

      sd=sd2,lun=/dev/dm-2

      sd=sd3,lun=/dev/dm-3

      sd=sd4,lun=/dev/dm-4

      sd=sd5,lun=/dev/dm-5

      sd=sd6,lun=/dev/dm-6

      sd=sd7,lun=/dev/dm-7

      sd=sd8,lun=/dev/dm-8

      sd=sd9,lun=/dev/dm-9

      sd=sd10,lun=/dev/dm-13

      sd=sd11,lun=/dev/dm-11

      sd=sd12,lun=/dev/dm-12

      sd=sd13,lun=/dev/dm-14

      sd=sd14,lun=/dev/dm-15

      sd=sd15,lun=/dev/dm-16

      wd=default,rdpct=0,xfersize=(4k,12,8k,17,16k,21,32k,30,64k,20),sd=sd*

      wd=65read_35write_write1,range=(22.5000,25.0000),skew=10

      wd=65read_35write_write2,range=(47.5000,50.0000),skew=25

      wd=65read_35write_write3,range=(72.5000,75.0000),skew=55

      wd=65read_35write_write4,range=(97.5000,100.0000),skew=10

      rd=run-def,iorate=10000,interval=60,elapsed=3600,pause=0,maxdata=999t,wd=*

       

       

       

       

      Here is how the failure looks like (with 10K IOPS)

      15:29:48.491

      15:29:48.494 Adjusted default JVM count for host=localhost from jvms=1 to jvms=2 because of iorate=10000.0 and a total of 16 sds.

      15:29:48.496

      15:29:48.560 Starting slave: /home/os76/lp-fb33-5/workspace/perf_benchmark/vdbench/vdbench SlaveJvm -m localhost -n localhost-10-150617-15.29.48.389 -l localhost-0 -p 5570

      15:29:48.580 Starting slave: /home/os76/lp-fb33-5/workspace/perf_benchmark/vdbench/vdbench SlaveJvm -m localhost -n localhost-11-150617-15.29.48.389 -l localhost-1 -p 5570

      15:29:48.976 All slaves are now connected

      15:30:24.636 Waiting for slave synchronization: localhost-0

      15:30:54.668 Waiting for slave synchronization: localhost-0

      15:31:24.701 Waiting for slave synchronization: localhost-0

      15:31:54.734 Waiting for slave synchronization: localhost-0

      15:32:24.767 Waiting for slave synchronization: localhost-0

      15:32:54.800 Waiting for slave synchronization: localhost-0

      15:33:24.831 Waiting for slave synchronization: localhost-0

      15:33:54.860 Waiting for slave synchronization: localhost-0

      15:34:05.348 localhost-1 : 15:34:05.348 common.failure(): System.exit(-99)

      15:34:05.348 localhost-1 :     at Vdb.SD_entry.openAllSds(SD_entry.java:186)

      15:34:05.348 localhost-1 :     at Vdb.SlaveWorker.doRegularWorkload(SlaveWorker.java:191)

      15:34:05.348 localhost-1 :     at Vdb.SlaveWorker.run(SlaveWorker.java:131)

      15:34:05.944

      15:34:05.944 Receiving unexpected EOFException from slave: localhost-1

      15:34:05.944 This means that this slave terminated prematurely.

      15:34:05.944 This thread will go to sleep for 5 seconds to allow

      15:34:05.944 slave termination to be properly recognized.

      15:34:05.944

      15:34:06.445

      15:34:06.445 Slave localhost-1 prematurely terminated.

      15:34:06.445 Look at file localhost-1.stdout.html for more information.

      15:34:06.446

      15:34:06.446 Slave localhost-1 prematurely terminated.

      15:34:06.446

      java.lang.RuntimeException: Slave localhost-1 prematurely terminated.

          at Vdb.common.failure(common.java:306)

          at Vdb.SlaveStarter.startSlave(SlaveStarter.java:179)

          at Vdb.SlaveStarter.run(SlaveStarter.java:50)

       

       

       

      Here is my java version:

      java version "1.7.0_75"

      OpenJDK Runtime Environment (rhel-2.5.4.0.el6_6-x86_64 u75-b13)

      OpenJDK 64-Bit Server VM (build 24.75-b04, mixed mode)

       

      Running on "CentOS release 6.5 (Final)"