From sliu at pipeline.com Mon May 12 17:11:02 2003 From: sliu at pipeline.com (Shaohui Liu) Date: Tue Nov 9 01:14:27 2010 Subject: [scyld-users] compilation errors on High Performance Computing Linpack Benchmark (HPL) Message-ID: <7894239.1052774079561.JavaMail.nobody@wamui02.slb.atl.earthlink.net> Hi, I am a new beowulf user. I am trying to bench mark my cluster which was built on top of some old PCs. My master is a PII, and the rest are PIs. Here is OS: Linux beowulf1 2.2.19-12.beo #1 Tue Jul 17 17:10:45 EDT 2001 i686 unknown. I was running a make program for HPL, and got an error which I could not solve. Here is the portion of the error msg ( I add echo to options): mpif77 -fomit-frame-pointer -O3 -funroll-loops -echo -o /home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a /usr/lib/libmpich.a + [ 0 = 1 ] + allargs= -fomit-frame-pointer -O3 -funroll-loops -o + [ 0 = 1 ] + linkargs= -fomit-frame-pointer -O3 -funroll-loops -o + compileargs= -fomit-frame-pointer -O3 -funroll-loops -c + gettinglinkarg=1 + DoLink=1 + [ 1 = 1 ] + linkargs= -fomit-frame-pointer -O3 -funroll-loops -o /home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl + gettinglinkarg=0 + outputfilename=/home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl + continue + [ 0 = 1 ] + allargs= -fomit-frame-pointer -O3 -funroll-loops -o HPL_pddriver.o + [ -s HPL_pddriver.o ] ++ expr HPL_pddriver.o : .*\(\..*\) + ext=.o + [ .o = .f -o .o = .F -o .o = .for -o .o = .FOR ] + [ .o = .o ] + [ 0 = 1 ] + DoLink=1 + linkobjs= HPL_pddriver.o + [ 0 = 1 ] + allargs= -fomit-frame-pointer -O3 -funroll-loops -o HPL_pddriver.o HPL_pdinfo.o + [ -s HPL_pdinfo.o ] ++ expr HPL_pdinfo.o : .*\(\..*\) + ext=.o + [ .o = .f -o .o = .F -o .o = .for -o .o = .FOR ] + [ .o = .o ] + [ 0 = 1 ] + DoLink=1 + linkobjs= HPL_pddriver.o HPL_pdinfo.o + [ 0 = 1 ] + allargs= -fomit-frame-pointer -O3 -funroll-loops -o HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o + [ -s HPL_pdtest.o ] ++ expr HPL_pdtest.o : .*\(\..*\) + ext=.o + [ .o = .f -o .o = .F -o .o = .for -o .o = .FOR ] + [ .o = .o ] + [ 0 = 1 ] + DoLink=1 + linkobjs= HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o + [ 0 = 1 ] + allargs= -fomit-frame-pointer -O3 -funroll-loops -o HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a + [ -s /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a ] ++ expr /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a : .*\(\..*\) + ext=.a + [ .a = .f -o .a = .F -o .a = .for -o .a = .FOR ] + [ .a = .o ] + compileargs= -fomit-frame-pointer -O3 -funroll-loops -c /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a + linkargs= -fomit-frame-pointer -O3 -funroll-loops -o /home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a + [ 0 = 1 ] + allargs= -fomit-frame-pointer -O3 -funroll-loops -o HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a + [ -s /usr/lib/libcblas.a ] ++ expr /usr/lib/libcblas.a : .*\(\..*\) + ext=.a + [ .a = .f -o .a = .F -o .a = .for -o .a = .FOR ] + [ .a = .o ] + compileargs= -fomit-frame-pointer -O3 -funroll-loops -c /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a + linkargs= -fomit-frame-pointer -O3 -funroll-loops -o /home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a + [ 0 = 1 ] + allargs= -fomit-frame-pointer -O3 -funroll-loops -o HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a + [ -s /usr/lib/libatlas.a ] ++ expr /usr/lib/libatlas.a : .*\(\..*\) + ext=.a + [ .a = .f -o .a = .F -o .a = .for -o .a = .FOR ] + [ .a = .o ] + compileargs= -fomit-frame-pointer -O3 -funroll-loops -c /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a + linkargs= -fomit-frame-pointer -O3 -funroll-loops -o /home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a + [ 0 = 1 ] + allargs= -fomit-frame-pointer -O3 -funroll-loops -o HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a /usr/lib/libmpich.a + [ -s /usr/lib/libmpich.a ] ++ expr /usr/lib/libmpich.a : .*\(\..*\) + ext=.a + [ .a = .f -o .a = .F -o .a = .for -o .a = .FOR ] + [ .a = .o ] + compileargs= -fomit-frame-pointer -O3 -funroll-loops -c /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a /usr/lib/libmpich.a + linkargs= -fomit-frame-pointer -O3 -funroll-loops -o /home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a /usr/lib/libmpich.a + added_link=0 + [ -z -I ] + FINC=-I/usr/include + status=0 + [ 0 = 1 -o 0 = 1 ] + [ 1 = 1 -o 0 = 1 ] + [ mpich = mpich ] + mpilibs=-lmpich + [ -n -o ! -s /usr/lib/libpmpich.a ] + [ yes = yes -a != yes ] + proflib= + [ -n ] + [ gcc != ignore -a no = yes -a -s /usr/lib/libmpichfarg.a ] + eval f77 -L/usr/lib HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o -fomit-frame-pointer -O3 -funroll-loops -o /home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a /usr/lib/libmpich.a -lmpich -lbproc -lpvfs -Wl,--undefined=beowulf_sched_shim,--undefined=get_beowulf_job_map -lbeomap -lbeostat -ldl ++ f77 -L/usr/lib HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o -fomit-frame-pointer -O3 -funroll-loops -o /home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a /usr/lib/libmpich.a -lmpich -lbproc -lpvfs -Wl,--undefined=beowulf_sched_shim,--undefined=get_beowulf_job_map -lbeomap -lbeostat -ldl /usr/bin/ld: cannot find -lpvfs collect2: ld returned 1 exit status + status=1 + [ 0 = 1 ] + exit 1 make[2]: *** [dexe.grd] Error 1 Does anyone know why? and how to solve it? Thanks a lot Shaohui From becker at scyld.com Mon May 12 18:09:01 2003 From: becker at scyld.com (Donald Becker) Date: Tue Nov 9 01:14:27 2010 Subject: [scyld-users] compilation errors on High Performance Computing Linpack Benchmark (HPL) In-Reply-To: <7894239.1052774079561.JavaMail.nobody@wamui02.slb.atl.earthlink.net> Message-ID: On Mon, 12 May 2003, Shaohui Liu wrote: > I am a new beowulf user. I am trying to bench mark my cluster which was built on top of some old PCs. My master is a PII, and the rest are PIs. > Here is OS: > Linux beowulf1 2.2.19-12.beo #1 Tue Jul 17 17:10:45 EDT 2001 i686 unknown. Ahhh, this is an old distribution from the 27 series, likely 27bz-9 or 27cz-9. > I was running a make program for HPL, and got an error which I could > not solve. Here is the portion of the error msg ( I add echo to > options): ... > ++ f77 -L/usr/lib HPL_pddriver.o HPL_pdinfo.o HPL_pdtest.o -fomit-frame-pointer -O3 -funroll-loops -o /home/sliu/hpl/bin/Linux_PII_CBLAS/xhpl /home/sliu/hpl/lib/Linux_PII_CBLAS/libhpl.a /usr/lib/libcblas.a /usr/lib/libatlas.a /usr/lib/libmpich.a -lmpich -lbproc -lpvfs -Wl,--undefined=beowulf_sched_shim,--undefined=get_beowulf_job_map -lbeomap -lbeostat -ldl > /usr/bin/ld: cannot find -lpvfs > collect2: ld returned 1 exit status Hmmm, if this is the low-cost basic edition (with "bz" in the part number), it did not ship with the PVFS file system libraries. The basic editions were intended as a minimal complexity, plug-and-go deployment platform. The PVFS (Parallel Virtual File System) implementation requires specific configuration to match the cluster configuration and application usage. Thus we provided the precompiled PVFS libraries with the professional and supported developer's releases, but not the basic/deployment releases. The solution is to -lpvfs from the compile line. -- Donald Becker becker@scyld.com Scyld Computing Corporation http://www.scyld.com 914 Bay Ridge Road, Suite 220 Scyld Beowulf cluster system Annapolis MD 21403 410-990-9993 From rgbose at monolith.wustl.edu Thu May 15 15:53:05 2003 From: rgbose at monolith.wustl.edu (Richard Bose) Date: Tue Nov 9 01:14:27 2010 Subject: [scyld-users] node problems Message-ID: I'm installing Scyld 28cz. The master is up and running. I made a node boot disk, the node booted off of it, the node showed up in beosetup, and reported status of error. Now I try to run beofdisk -n 0 -d and it says: cat: /lib/libc.so.6: version `GLIBC_2.2' not found (required by cat) No disks found on node 0. And it says something similar for bpsh -n 0 uptime: uptime: relocation error: /lib/libc.so.6: symbol _dl_debug_impcalls, version GLIBC_2.0 not defined in file ld-linux.so.2 with link time reference How can I resolve this? Thanks From becker at scyld.com Thu May 15 16:10:07 2003 From: becker at scyld.com (Donald Becker) Date: Tue Nov 9 01:14:27 2010 Subject: [scyld-users] node problems In-Reply-To: Message-ID: On Thu, 15 May 2003, Richard Bose wrote: > I'm installing Scyld 28cz. The master is up and running. I made a node > boot disk, the node booted off of it, the node showed up in beosetup, and > reported status of error. This usually means that it failed to mount a filesystem specified in /etc/beowulf/fstab > Now I try to run beofdisk -n 0 -d and it says: > cat: /lib/libc.so.6: version `GLIBC_2.2' not found (required by cat) ... > How can I resolve this? 1. Comment out the failing entry in /etc/beowulf/fstab 2. Run beofdisk 3. Replace the mount entries in /etc/beowulf/fstab Note that there is a method for having per-node fstab files. For node 23 the search order is /etc/beowulf/fstab.23 /etc/beowulf/fstab -- Donald Becker becker@scyld.com Scyld Computing Corporation http://www.scyld.com 914 Bay Ridge Road, Suite 220 Scyld Beowulf cluster system Annapolis MD 21403 410-990-9993 From sliu at pipeline.com Wed May 28 14:08:00 2003 From: sliu at pipeline.com (Shaohui Liu) Date: Tue Nov 9 01:14:27 2010 Subject: [scyld-users] beostat error Message-ID: <3ED4FF81.3040009@pipeline.com> Hi, I am running a cluster of 9 nodes with OS: Linux beowulf1 2.2.19-12.beo #1 Tue Jul 17 17:10:45 EDT 2001 i686 unknown I have been able to compile and run a benchmark program under parallel mode. With a command like bpsh -n 0 vmstat, I could see the benchmark is running under different nodes. But I could not see any change on BeoStatus monitoring window. Plus, the monitoring window only shows CPU and memory for three nodes. Running command beostat, I could get output for three nodes, which are the same nodes on the monitoring window. I have set export NP=10 and export ALL_CPUS=1 in login console, I did not see any change. Running beomap command shows -1:0:1:2:3:4:5:6:7 . Plus, by starting beostatus from login console, after a while, I could see some errors msg like: BeoStat library CPU count call failed: Too many open files Can anyone help me this? Thanks in advance Shaohui Liu