Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: Re: [OMPI users] OpenMPI Hangs, No Error
From: Robert Walters (raw19896_at_[hidden])
Date: 2010-07-14 11:28:45


Also, I finally got some graphical output from Sun Studio Analyzer.

I see MPI_Recv and MPI_Wait taking a lot of time, but I would think that is ok, this program does heavy number crunching and I would expect it to need to Wait or wait to Receive very often since there is a decent amount of time between communications. Is this the correct assumption?

What does catch my eye is MPI_Barrier takes up a significant chunk of around 10%. I read that MPI_Barrier blocks the caller until all processes have called? Perhaps there is something fishy there that it is taking an awful long time for processes to call each other although MPI_Send is not taking very long so it makes me feel more comfortable about network communication.

Anyways, please have a look and let me know what you think could be the issue.

Regards,
Robert Walters

--- On Tue, 7/13/10, Robert Walters <raw19896_at_[hidden]> wrote:

From: Robert Walters <raw19896_at_[hidden]>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error
To: "Open MPI Users" <users_at_[hidden]>
Date: Tuesday, July 13, 2010, 10:42 PM

Naturally, a forgotten attachment.

An to edit that, it was compiled to be used with OpenMPI 1.4.1, but as I understand, 1.4.2 is just a bug fix of 1.4.1.

--- On Tue, 7/13/10, Robert Walters <raw19896_at_[hidden]> wrote:

From: Robert Walters <raw19896_at_[hidden]>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error
To: "Open MPI Users" <users_at_[hidden]>
Date: Tuesday, July 13, 2010, 10:38 PM

I think I forgot to mention earlier that the application I am using is pre-compiled. It is a finite element software called LS-DYNA. It is not open source and I likely cannot obtain the code it uses for MPP. This version I am using was specifically compiled, by the parent company, for OpenMPI 1.4.2 MPP operations.

I recently installed the Sun Studio 12.1 to attempt to analyze the situation. It seems to work partially. It will record various processes individually, which is cryptic. The function it fails on, though, is the MPI Tracing. It errors that "no MPI tracing data file in experiment, MPI Timeline and MPI Charts will not be available". Sometime during the analysis (about 10,000 iterations later, the VT_MAX_FLUSHES complains that there are too many i/o flushes and its not happy. I've increased this number in the environmental variable and killed the analysis before it had a chance to error but still no MPI Trace data is recorded. Not sure if you guys have heard of that happening or know any way to fix it...Did OpenMPI need to be configured/built for Sun Studio use?

I also noticed that from the data I do get back, there are two sets of functions for everything. There is mpi_recv and then my_recv_, both with the same % utilization time. The mpi one comes from your program's library and the my_recv_ one comes from my program. Is that typical or should the program I'm using be saying mpi_recv only? This data may be enough to help me see what's wrong so I will pass it along. Keep in mind this is percent time of total run time and not percent of MPI communication. I attached the information in a picture rather than me attempting to format a nice table in this nasty e-mail application.
I blacked out items that are related to LS-DYNA but afterward I just realized that I think every function with an _ at the end represents a command issuing from LS-DYNA.

These are my big spenders. The processes I did not include are in the bottom 4%. The processes that would be above these were the LS-DYNA applications at 100%. Like I mentioned earlier, there are two instances of every MPI command, and they carry the same percent usage. It's curious that this version, built for OpenMPI, uses different functions.

Just for a little more background info, OpenMPI is being launched from a local hard drive on each machine, but the LS-DYNA job files, and related data output files, are on a mounted drive on that machine, where the mounted drive is located on a different machine also in the cluster. We were thinking that might be an issue but it isn't writing enough data for me to think that would significantly decrease MPP performance.

I would like to make one last mention. That is that OpenMPI running 8 cores on a single node, with all the communication, works flawlessly. It works much faster than the Shared Memory Parallel (SMP) version of LS-DYNA that we currently have used scaled to 8 cores. LS-DYNA seems to be approximately 25% faster (don't quote me on that) when using the OpenMPI installation than when using the standard SMP, which is awesome. My point being that OpenMPI seems to be working fine, even with the screwy mounted drive. This leads me to continue to point at the network.

Anyhow, let me know if anything seems weird on the OpenMPI communication subroutines. I don't have any numbers to lean on from experience.

Sorry this e-mail was long. Thank you again for all of your help.

Regards,
Robert Walters

--- On Tue, 7/13/10, David Zhang <solarbikedz_at_[hidden]> wrote:

From: David Zhang <solarbikedz_at_[hidden]>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error
To: "Open MPI Users" <users_at_[hidden]>
Date: Tuesday, July 13, 2010, 9:42 PM

Like Ralph says, the slow down may not be coming from the kernel, but rather on waiting for messages.  What MPI send/recv commands are you using?

On Tue, Jul 13, 2010 at 11:53 AM, Ralph Castain <rhc_at_[hidden]> wrote:

The most likely thing that is happening is that OMPI is polling waiting for messages to arrive. You might look closer at your code to try and optimize it better so that number-crunching can get more attention.
Others on this list are far more knowledgeable than I am about doing such things, so I'll let them take it from here. Glad it is now running!

On Jul 13, 2010, at 12:22 PM, Robert Walters wrote:
OpenMPI,

Following up. The sysadmin opened ports for machine to machine communication and OpenMPI is running successfully with no errors in connectivity_c, hello_c, or ring_c. Since, I have started to implement our MPP software (finite element analysis) that we have, and upon running a simple, 1 core on machine1, 1 core on machine2, job, I notice it is considerably slower than a 2 core job on a single machine.

A quick look at top shows me kernel usage is almost twice what cpu usage is! On a 16 core job, (8 cores per node so 2 nodes total) test, OpenMPI was consuming ~65% of the cpu for kernel related items rather than number-crunching related items...Granted, we are running on GigE, but this is a finite element code we are running with no heavy data transfer within it. I'm looking into benchmarking tools, but my sysadmin is not very open to installing third party softwares. Do you have any suggestions for what I can use that would be "big name" or guaranteed safe tools I can use to figure out what's causing the hold up with all the kernel usage? I'm pretty sure its network traffic but I have no way of telling (as far as I know because I'm not a Linux whiz) with the standard tools in RHEL.

Thanks for all the help! I'm glad to get it finally working and I think with a little tweaking it should be ready to go very soon.

Regards,
Robert Walters
--- On Sat, 7/10/10, Ralph Castain <rhc_at_[hidden]> wrote:

From: Ralph Castain <rhc_at_[hidden]>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error

Date: Saturday, July 10, 2010, 4:37 PM

The "static ports" flag means something different - it is used when the daemon is given a fixed port to use. In some installations, we lock every daemon to the same port number so that each daemon can compute exactly how to contact its peers (i.e., no contact info exchange required for wireup).
You have a "fixed range", but not "static port", scenario. Hence the message.
Let us know how it goes - I agree it sounds like something to discuss with the sysadmin.

On Jul 10, 2010, at 1:47 PM, Robert Walters wrote:
I ran oob_tcp_verbose 99 and I am getting something interesting I never got before.

[machine 2:22347] bind() failed: no port available in the range [60001-60016]
[machine 2:22347] mca_oob_tcp_init: unable to create IPv4 listen socket: Error

I never got that error before we messed with the iptables but now I get that error... Very interesting, I will have to talk to my sysadmin again and make sure he opened the right ports on my two test machines. It looks as though there are no open ports. Another interesting thing is I see that the Daemon is still report:

Daemon [[28845,0],1] checking in as pid 22347 on host machine 2
Daemon [[28845,0],1] not using static ports

Which, I may be misunderstanding, should have been taken care of when I specified what ports to use. I am telling it a static set of ports... Anyhow, I will get with my sysadmin again and see what he says. At least OpenMPI is correctly interpreting the range.

Thanks for the help.

--- On Sat, 7/10/10, Ralph Castain <rhc_at_[hidden]> wrote:

From: Ralph Castain <rhc_at_[hidden]>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error
To: "Open MPI Users" <users_at_[hidden]>

Are there multiple interfaces on your nodes? I'm wondering if we are using a different network than the one where you opened these ports.
You'll get quite a bit of output, but you can turn on debug output in the oob itself with -mca oob_tcp_verbose xx. The higher the number, the more you get.

On Jul 10, 2010, at 11:14 AM, Robert Walters wrote:
Hello again,

I believe my administrator has opened the ports I requested. The problem I am having now is that OpenMPI is not listening to my defined port assignments in openmpi-mca-params.conf (looks like permission 644 on those files should it be 755?)

When I perform netstat -ltnup I see that orted is listening 14 processes in tcp but scaterred in the 26000ish port range when I specified 60001-60016 in the mca-params file. Is there a parameter I am missing? In any case I am still hanging as mentioned originally even with the port forwarding enabled and specifications in mca-param enabled.

Any other ideas on what might be causing the hang? Is there a more verbose mode I can employ to see more deeply into the issue? I have run --debug-daemons and --mca plm_base_verbose 99.

Thanks!
--- On Tue, 7/6/10, Robert Walters <raw19896_at_[hidden]> wrote:

From: Robert Walters <raw19896_at_[hidden]>

To: "Open MPI Users" <users_at_[hidden]>
Date: Tuesday, July 6, 2010, 5:41 PM

Just to confirm with you, I should change openmpi-mca-params.conf to include:

oob_tcp_port_min_v4 = (My minimum port in the range)
oob_tcp_port_range_v4 = (My port range)
btl_tcp_port_min_v4 = (My minimum port in the range)
btl_tcp_port_range_v4 = (My port range)

correct?

Also, for a cluster of around 32-64 processes (8 processors per node), how wide of a range will I require? I've noticed some entries in the mailing list suggesting you need a few to get started and then it opens as necessary. Will I be safe with 20 or should I go for 100?

Thanks again for all of your help!

--- On Tue, 7/6/10, Ralph Castain <rhc_at_[hidden]> wrote:

From: Ralph Castain <rhc_at_[hidden]>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error
To: "Open MPI Users" <users_at_[hidden]>
Date: Tuesday, July 6, 2010, 5:31 PM

Problem isn't with ssh - the problem is that the daemons need to open a TCP connection back to the machine where mpirun is running. If the firewall blocks that connection, then we can't run.

Yes, there is a system firewall. I don't think the sysadmin will allow it to go disabled. Each Linux machine has the built-in RHEL firewall. SSH is enabled through the firewall though.

--- On Tue, 7/6/10, Ralph Castain <rhc_at_[hidden]> wrote:

From: Ralph Castain <rhc_at_[hidden]>
Subject: Re: [OMPI users] OpenMPI Hangs, No Error
To: "Open MPI Users" <users_at_[hidden]>
Date: Tuesday, July 6, 2010, 4:19 PM

It looks like the remote daemon is starting - is there a firewall in the way?
On Jul 6, 2010, at 2:04 PM, Robert Walters wrote:
Hello all,

I am using OpenMPI 1.4.2 on RHEL. I have a cluster of AMD Opteron's and right now I am just working on getting OpenMPI itself up and running. I have a successful configure and make all install. LD_LIBRARY_PATH and PATH variables were correctly edited. mpirun -np 8 hello_c successfully works on all machines. I have setup my two test machines with DSA key pairs that successfully work with each other.

The problem comes when I initiate my hostfile to attempt to communicate across machines. The hostfile is setup correctly with <host_name> <slots> <max-slots>. When running with all verbose options enabled "mpirun --mca plm_base_verbose 99 --debug-daemons --mca btl_base_verbose 30 --mca oob_base_verbose 99 --mca pml_base_verbose 99 -hostfile hostfile -np 16 hello_c" I receive the following text output.

[machine1:03578] mca: base: components_open: Looking for plm components
[machine1:03578] mca: base: components_open: opening plm components
[machine1:03578] mca: base: components_open: found loaded component rsh

[machine1:03578] mca: base: components_open: component rsh open function successful
[machine1:03578] mca: base: components_open: found loaded component slurm

[machine1:03578] mca: base: components_open: component slurm open function successful
[machine1:03578] mca:base:select: Auto-selecting plm components

[machine1:03578] mca:base:select:(  plm) Query of component [rsh] set priority to 10
[machine1:03578] mca:base:select:(  plm) Querying component [slurm]
[machine1:03578] mca:base:select:(  plm) Skipping component [slurm]. Query failed to return a module
[machine1:03578] mca:base:select:(  plm) Selected component [rsh]
[machine1:03578] mca: base: close: component slurm closed

[machine1:03578] mca: base: components_open: Looking for oob components
[machine1:03578] mca: base: components_open: opening oob components
[machine1:03578] mca: base: components_open: found loaded component tcp

[machine1:03578] mca: base: components_open: component tcp open function successful
Daemon was launched on machine2- beginning to initialize

[machine2:01962] mca: base: components_open: opening oob components
[machine2:01962] mca: base: components_open: found loaded component tcp
[machine2:01962] mca: base: components_open: component tcp has no register function
[machine2:01962] mca: base: components_open: component tcp open function successful
Daemon [[1418,0],1] checking in as pid 1962 on host machine2

At this point the system hangs indefinitely. While running top on the machine2 terminal, I see several things come up briefly. These items are: sshd (root), tcsh (myuser), orted (myuser), and mcstransd (root). I was wondering if sshd needs to be initiated by myuser? It is currently turned off in sshd_config through UsePAM yes. This was setup by the sysadmin but it can be worked around if this is necessary.

So in summary, mpirun works on each machine individually, but hangs when initiated through a hostfile or with the -host flag. ./configure with defaults and --prefix. LD_LIBRARY_PATH and PATH set up correctly. Any help is appreciated. Thanks!

users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users

-----Inline Attachment Follows-----

_______________________________________________
users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users

users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users

-----Inline Attachment Follows-----

_______________________________________________
users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users

users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users

-----Inline Attachment Follows-----

_______________________________________________
users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users

users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users

-----Inline Attachment Follows-----

_______________________________________________
users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users

users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users

_______________________________________________

-- 
David Zhang
University of California, San Diego
-----Inline Attachment Follows-----
_______________________________________________
users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users
-----Inline Attachment Follows-----
_______________________________________________
users mailing list
users_at_[hidden]
http://www.open-mpi.org/mailman/listinfo.cgi/users



sungraph.jpg