---+ *Migrate on the SCS Condor cluster (Vanilla universe)*   

1) log in one of the SCS master nodes

SCS has several master nodes in the the Condor Pool. You may or may not be eligible for access to all of them. Master nodes available for general use include phoenix000 (Phoenix Cluster), tempest01 (Tempest Cluster), and prism01 (Prism Cluster). The following nodes require permission from their respective owners: anfinsen098 (Anfinsen Cluster), petal017 (Petal Cluster), grove000 (Grove Cluster), and waterfall000 (TechHelp?.Waterfall Cluster). For more information, please refer to the SCS High Performance Clusters .

So you can log in the general use clusters like phoenix, tempest and prism and relink it by yourself at your dirctory.

ssh username@phoenix.csit.fsu

2) Install Migrate at your directory

The Migrate hasn't been installed at /cluster/bin/ directory. So before you run your job on condor vanilla universe, you need to install it to at your directory. You can find how to install the Migrate documentation downloadable at Dr. Beerli's Migrate website . Or you can follow me step by step.

a) Download the source accoring your operating system at Dr. Beerli's Migrate website

b) Suppose we use Linux(intel) operating system, unzip it with gunzip -c migrate-2.0.3.src.tar.gz | tar xf - or use tar xvfz migrate-2.0.3.src.tar.gz [this creates a directory "migrate-2.0.3" with subdirectories "src", and "examples" in it.]

c) cd migrate-2.0.3/src

d) type "./configure". This will create the Makefile

e) use "make" command to install it.(It took about 1 minutes on my computer)

f) use command "ls -l" to check check your dirctory and you will find an executable file "migrate-n" there.

g) run it with "./migrate" command and you will see something like below.

 
[yanfeng@phoenix000 src]$ ./migrate-n
=============================================
  MIGRATION RATE AND POPULATION SIZE ESTIMATION
  using Markov Chain Monte Carlo simulation
  =============================================
  Version  2.0.6
  Program started at   Mon Jun  6 15:52:01 2005


  Settings for this run:
  D       Data type currently set to:             DNA sequence model
  I       Input/Output formats
  P       Parameters  [start, migration model]
  S       Search strategy
  W       Write a parmfile
  Q       Quit the program


  Are the settings correct?
  (Type Y or the letter for one to change) 
 

To leave the program, just type quit or q and you should get the Unix prompt back.

h) It is not convient to use the migrate program at this dirctory. So you can copy it to a directory easy to find or create a link for that in the directory easy to find. At my home directory, I created a dirctory called "condor_dir" and in that directory I created a directory called "mig". I copy the migrate program there with a new name "local_mig" so that it will reminder me it is local Migrate(or we can say local version Migrate, not the one relinked against the condor library used for standard universe). T0 do that, return to the directory where the Migrate is. Then use command "cp migrate-n ~/condor_dir/mig/local_mig" to copy it to your "mig" directory.

i) Or you can just create a link for your relinked Migrate. Use command "ln -s migrate-n ~/condor_dir/mig/local_mig" to creat a hard link at your "mig" directory.(use "man ln" to know more ln command)

3) prepare your Migrate input file (and parmfile)

Before we start pratice Migrate on condor, it is usful to set up a directory to work in it. You need to konw three simple terminal commands: mkdir, cd ,and cp. If you have program with them, use man mkdir, man cd, man cp to study it fist.

Now creat a directory called condor_dir in your home directory. Go to that directory and creat another directory called mig. (omit this step if you have done that above.)

There are several different input data type, such as electrophoretic marker data, microsatellite data, sequence data. We first practice on sequence data. You can find the data in the Migrate package example directory or you download below. Save it to your mig directory. Go to mig directory and create a new directoy call mig_1_vanilla. Copy the example data file to mig_1_vanilla directoy. To see the contents of the file, use command " cat infile.seq ".

 
   4 2  Example: sequence data set wit two loci [simulated data]
145 345
25   Africa
0BAA      AAAGCTTTGGAAAAACATTGGACGCAAGGATACGG ......
0BAN      AAAGCTTTGGAAAAACATTCGAGGCAAGGATACGG ......
 ......
 
25   Americas
1BAX      AAAGCTTTGGAAAAACATTCGAGGCAAGGATACGG ......
1BAK      AAAGCTTTGGAAAAACATTCGAGGCAAGGATACGG ......
  ......
 
25   Pacific
2BAP      AAAGCTTTGGAAAAACATTGGACGCAAGGATACGG ......
2BAQ      AAAGCTTTGGAAAAACATTGGACGCAAGGATACGG ......
  
 ......
 
25   Asia
3BAH      AAAGCTTTGGAAAAACATTGGACGCAAGGATACG ......
3BAR      AAAGCTTTGGAAAAACATTGGACGCAAGGATACG ......
 
 ......
 

download infile.seq here

I would not explain the data file form here. Please read the chapter "Data file specification" in the documetation downloadable at Dr. Beerli's websit .

Except for the input data file, there are some other files may be used in the anlysis, like parmfile, geofile and seedfile. The parmfile is especially worth to be mentioned. Parmfile can hold specific menu options. Without it, Migrate will display a menu, in which you can change all the sensible options. But with parmfile, you can edit the parmfile to choose the options than making the changes in the menu. It is very like the MrBayes batch block and PAUP batch block in the Nexus file.

Then what does parmfile look like? There are several parmfiles in the Migrate packege or you can download below. Use command " cat parmfile.seq " to view it.

 
#########################################################
# Parmfile for Migrate 0.9.1
# generated automatically on
# 04/13/00 12:39:45
#
# please report problems to Peter beerli
#    email: beerli@genetics.washington.edu
#    http://evolution.genetics.washington.edu/lamarc.html
#########################################################
# General options ---------------------------------------
nmlength=10
# data options ------------------------------------------
datatype=SequenceData
# Sequence data options------------
ttratio=0.600000 15.000000 
freqs-from-data=YES
categories=1
rates=1:1.000000 
prob-rates=1:1.000000 
autocorrelation=NO
weights=NO
interleaved=NO
usertree=NO
distfile=NO
# input/output options ----------------------------------
menu=NO # Change YES to NO
# input formats
infile=infile.seq
random-seed=AUTO #OWN:955654772
# output formats
progress=YES
print-data=NO
outfile=outfile.seq
plot=NO
profile=ALL:FAST
print-tree=NONE
mathfile=mathfile
write-summary=NO
# likelihood-ratio test
# parameter options ------------------------------------
theta=FST
migration=FST
mutation=NOGAMMA
fst-type=THETA
custom-migration={**}
# search strategies ------------------------------------
short-chains=10
short-inc=20
short-sample=500
long-chains=3
long-inc=20
long-sample=5000
#obscure options
burn-in=10000
heating=NOmoving-steps=NO
long-chain-epsilon=100.000000
gelman-convergence=No
replicate=Yes:LastChains
end
 

download parmfile.seq here

For the meaning of the options, please refer to the chapter "Menu and Options" in the documetation downloadable at Dr. Beerli's websit . I can't explain better than that. One thing you need to make sure is set "menu" to "NO" such that the program will not show up the menu and run it directlly accordding to the options you set in the parmfile directly. The default is "YES". Another thing you can't omit is to set the infile to your data file name such that the program can find your data file. The default name is "infile".(Or you can change your data file name to "infile")

There are two ways you can get the parmfile.The first way is edit it according to example parmfiles in the program package. The second way, which is also what I prefer to, is to run the Mirgate(local version or condor version) and set (1) Data type (2) Input/Output formats (3) Parameters (4) Search strategy as what you want. Then choose the option "W" (write a parmfile) and you will see "+++ Parmfile written to current directory +++" on the screen. Quit Migrate with "Q" , check your directory with "ls" command and you will find a file called "parmfile" created. (If you already have a file named "parmfile", it will be overwrited.) There is more detailed explaination for the options in the parmfile than the example parmfile above.( The parmfile from download link is created by running migrate)

Before you submit your jobs on condor, I would suggest you run it locally("./local_mig parmfile" ) first to see if it crashs immediately or soon after you run it. Make sure it can run correctly for a while, then quit with ctrl+C and delete all the files the program just created. After that, submit your job to condor.

4) creat your submit description file

Now you can create a condor submit description file. If you have general questions about how to create a condor submit description file, please refer to the General Program Tutorial .

 
########################################
#
# a simple Migrate job on condor vanilla universe
#
#########################################
InitialDir   = /home/u5/users/yanfeng/condor_dir/mig/mig_1_vanilla/
Universe     = vanilla
Executable   = /home/u5/users/yanfeng/condor_dir/mig/local_mig

Arguments    = parmfile.seq
 

output       = condor_mig_seq.output
error        = condor_mig_seq.error 
log          = condor_mig_seq.log

Should_transfer_files = YES
when_to_transfer_output = ON_EXIT_OR_EVICT
transfer_input_files = infile.seq, parmfile.seq
Queue

 

download submit description file mig_1_vanilla.submit

Tips

Under Unix, the Condor presumes a shared file system for vanilla jobs. So normally you donot need to specify the Initialdir. But the automounter on some condor nodes does not like how condor is referencing the files on it. This is due to how your shell's pwd command returns pathing. Some pwd implementations display /home... and some will display /a... This is dependent on your shell. Forcing condor to reference an initialdir value of /home/...... will make the automounter behave correctly.

Take my account for example, With "pwd" command, I find my program file and submit description file are locate at "/a/fs/u5/users/yanfeng/condor_dir/mig/mig_1_vanilla". I write the initialdir attribute in your submit file to"/home/u5/users/yanfeng/condor_dir/mig/mig_1_vanilla ".

And if you have file need to transfer, you'd better add use File Tranfer Mechanism.

   should_transfer_files   = YES
   when_to_transfer_output = ON_EXIT_OR_EVICT 
   transfer_input_files = file1,file2

5) submit your job and manage it

Use condor_submit command to submit your job and use condor_q, condor_rm etc to manage your job.

[yanfeng@phoenix000 mig_1_vanilla]$ condor_submit mig_1_vanilla.submit

 
 
Submitting job(s).
Logging submit event(s).
1 job(s) submitted to cluster 228. 
  

[yanfeng@phoenix000 mig_1_vanilla]$ condor_q yanfeng

  
-- Submitter: phoenix000.csit.fsu.edu : <144.174.160.169:9989> : phoenix000.csit.fsu.edu
 ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD   
 228.0   yanfeng         6/6  16:14   0+00:00:00 I  0   0.6  local_mig parmfile

1 jobs; 1 idle, 0 running, 0 held 
  

[yanfeng@phoenix000 mig_1_vanilla]$ ls -l

 
 
total 156   
-rw-r--r--    1 yanfeng  student         0 Jun  6 16:20 condor_mig_seq.error [condor error file]
-rw-r--r--    1 yanfeng  student       695 Jun  6 16:20 condor_mig_seq.log [condor log file]
-rw-r--r--    1 yanfeng  student     11418 Jun  6 16:20 condor_mig_seq.output [condor output file]
-rwxr--r--    1 yanfeng  student     51322 Jun  3 20:24 infile.seq [migrate infile]
-rw-------    1 yanfeng  student      8293 Jun  6 16:20 logfile.seq [migrate logfile]
-rw-r--r--    1 yanfeng  student       558 Jun  6 16:04 mig_1_vanilla.submit  [condor submit file]
-rw-------    1 yanfeng  student     62457 Jun  6 16:20 outfile.seq  [migrate outfile]
-rw-r--r--    1 yanfeng  student     19908 Jun  3 20:24 parmfile.seq [migrate parmfile]
 
 

You can open the log file to check the status of your job. cat condor_mig_seq.log

 
000 (228.000.000) 06/06 16:14:52 Job submitted from host: <144.174.160.169:9989>
...
001 (228.000.000) 06/06 16:17:02 Job executing on host: <144.174.160.206:9706>
...
006 (228.000.000) 06/06 16:17:10 Image size of job updated: 3112
...
005 (228.000.000) 06/06 16:21:04 Job terminated.
        (1) Normal termination (return value 0)
                Usr 0 00:03:46, Sys 0 00:00:00  -  Run Remote Usage
                Usr 0 00:00:00, Sys 0 00:00:00  -  Run Local Usage
                Usr 0 00:03:46, Sys 0 00:00:00  -  Total Remote Usage
                Usr 0 00:00:00, Sys 0 00:00:00  -  Total Local Usage
        82168  -  Run Bytes Sent By Job
        649394  -  Run Bytes Received By Job
        82168  -  Total Bytes Sent By Job
        649394  -  Total Bytes Received By Job
...

  

It means that condor finished your job without error.

6) more simple examples

example, run two Migrate jobs

Go to mig directory and creat another directory called mig_2_vanilla. Copy the example data file "infile.seq" to the new directory. Suppose we want to change the random number seed to check if we can get similar results with different seeds. You can use the parmfile.seq above. Use "mv" and "cp" command to create two copies of parmfile with name parmfile.seq.1 and parmfile.seq.2(" cp parmfile.seq parmfile.seq.1 " and " mv parmfile.seq.1 parmfile.seq.2 "). Open them with vi or other editors. In the parmfile.seq.1 file, change " "random-seed=AUTO" to "random-seed=OWN:279454829" and in the parmfile.seq.2 file, change "random-seed=AUTO" to "random-seed=OWN:279454825". ( random seedvalue can be any number but n*4+1 is pefered).Also change the output file name from "outfile.seq" to "outfile.seq.1" and change the log file name form "logfile.seq" to "logfile.seq.1" in the parmfile.seq.1. Simlar changes to parmfile.seq.2 to avoid the overwritting.

download here parmfile.seq.1

download here parmfile.seq.2

download here infile.seq

2) Your submit description file should like this (file name mig_2_vanilla.submit)

 

########################################
#
# 2 simple Migrate job on condor vanilla universe
#
#########################################
InitialDir   = /home/u5/users/yanfeng/condor_dir/mig/mig_2_vanilla/
Universe     = vanilla
Executable   = /home/u5/users/yanfeng/condor_dir/mig/local_mig

output       = condor_mig_seq.output.$(Process)
error        = condor_mig_seq.error.$(Process)
log          = condor_mig_seq.log

Should_transfer_files = YES
when_to_transfer_output = ON_EXIT_OR_EVICT

Arguments = parmfile.seq.1
transfer_input_files = infile.seq, parmfile.seq.1
Queue

Arguments = parmfile.seq.2
transfer_input_files = infile.seq, parmfile.seq.2
Queue

 

download here mig_2_vanilla.submit

3) You can use condor_q yourusername to check the stauts of your job. After it is finished, use ls -l to check your files in your directory.

[yanfeng@phoenix000 mig_2_vanilla]$ condor_q yanfeng

 
 
-- Submitter: phoenix000.csit.fsu.edu : <144.174.160.169:9989> : phoenix000.csit.fsu.edu
 ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD 
 230.0   yanfeng         6/6  16:21   0+00:01:03 R  0   0.6  local_mig parmfile
 230.1   yanfeng         6/6  16:21   0+00:00:00 I  0   0.6  local_mig parmfile
2 jobs; 0 idle, 2 running, 0 held
 

[yanfeng@phoenix000 mig_2_vanilla]$ ls -l

 
 
total 248  
-rw-r--r--    1 yanfeng  student         0 Jun  6 16:29 condor_mig_seq.error
-rw-r--r--    1 yanfeng  student      1392 Jun  6 16:29 condor_mig_seq.log
-rw-r--r--    1 yanfeng  student     11412 Jun  6 16:29 condor_mig_seq.output
-rwxr--r--    1 yanfeng  student     51322 Jun  3 20:30 infile.seq
-rw-------    1 yanfeng  student      8287 Jun  6 16:25 logfile.seq.1
-rw-------    1 yanfeng  student      8287 Jun  6 16:29 logfile.seq.2
-rw-r--r--    1 yanfeng  student       642 Jun  6 16:20 mig_2_vanilla.submit
-rw-------    1 yanfeng  student     62452 Jun  6 16:25 outfile.seq.1
-rw-------    1 yanfeng  student     62312 Jun  6 16:29 outfile.seq.2
-rw-r--r--    1 yanfeng  student     19906 Jun  3 20:30 parmfile.seq.1
-rw-r--r--    1 yanfeng  student     19906 Jun  3 20:30 parmfile.seq.2 

 

Tips

Sometime you may want the same outfile and logfile name for your two condor jobs. Then you need to create two directory so that they will not overwrite each other.Create two directory called "parm1" and "parm2". Copy your parmfiles(parmfile.seq.1 and parmfile.seq.2) to the directories. And also copy infile to the directories repectively. Change back the output file name(outfile.seq) and log file name(logfile.seq) in both parmfiles. But still keep the seeds to be different.(Becaue the parmfile names are conflict with the parmfile names of the example above when I uploaded the files to this webpage, I changed them to "parmfile.seq.a" and "parmfile.seq.b".)

-----|
     |- mig_2_vanilla_same.submit (file)
     |- parm1 (folder)
     |  |- infile.seq (file)
     |  |- parmfile.seq.a (file)
     |    
     |- parm2 (folder)
        |- infile.seq (file)
        |- parmfile.seq.b (file)

download here parmfile.seq.a

download here parmfile.seq.b

download here infile.seq

Then the submit description file will be something like below (file name mig_2_vanilla_same.submit).

 

########################################
#
# 2 simple Migrate job on condor vanilla universe
#
#########################################
Universe     = vanilla
Executable   = /home/u5/users/yanfeng/condor_dir/mig/local_mig

output       = condor_mig_seq.output
error        = condor_mig_seq.error 
log          = condor_mig_seq.log

Should_transfer_files = YES
when_to_transfer_output = ON_EXIT_OR_EVICT

InitialDir   = /home/u5/users/yanfeng/condor_dir/mig/mig_2_vanilla/two/parm1
Arguments = parmfile.seq.a
transfer_input_files = infile.seq, parmfile.seq.a
Queue

InitialDir   = /home/u5/users/yanfeng/condor_dir/mig/mig_2_vanilla/two/parm2
Arguments = parmfile.seq.b
transfer_input_files = infile.seq, parmfile.seq.b
Queue

 

download here mig_2_vanilla_same.submit

After the jobs are finished, you will find the files like below.

-----|
     |- mig_2_vanilla_same.submit (file)
     |- parm1 (folder)
     |  |- condor_mig_seq.error  (file)
     |  |- condor_mig_seq.log (file)
     |  |- condor_mig_seq.output (file)
     |  |- infile.seq (file)
     |  |- logfile.seq (file)
     |  |- outfile.seq (file)
     |  |- parmfile.seq.a (file)
     |  
     |- parm2 (folder)
        |- condor_mig_seq.error (file)
        |- condor_mig_seq.log (file)
        |- condor_mig_seq.output (file)
        |- infile.seq (file)
        |- logfile.seq (file)
        |- outfile.seq (file)
        |- parmfile.seq.b (file)

7) about parallel migrate.

At the page 60 of the documentation of migrate, there is toturial on how to run parallel migrate. One way is to use the standard message passing interface(MPI). We will not discuss it now. The other way is by hand. Instead of secure computers for the analysis, we can use condor to run seveal jobs on condor. It is very similar to the example above( run two jobs). You can practice by yourself.

-- YanfengShi - 09 May 2005