[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

19990916: McIDAS: Disk Partitioning

>From: Michael Keables <address@hidden>
>Organization: DU
>Keywords: 199909152035.OAA16060 LDM ldm-mcidas


>First, thanks so much for all of your help with cyclone and the LDM. Being
>a unix novice, I make a lot of mistakes and as such, I really appreciate
>not only your assistance but the explanations as well.

After I got your voicemail, we ran amok on cyclone :-)

>Here's the deal with cyclone. It has two hard drives installed in it with
>/export/home residing on one and everything else on the other:
> df -k
>Filesystem            kbytes    used   avail capacity  Mounted on
>/proc                      0       0       0     0%    /proc
>/dev/dsk/c0t0d0s0      96455   40768   46042    47%    /
>/dev/dsk/c0t0d0s6     877790  665487  150858    82%    /usr
>fd                         0       0       0     0%    /dev/fd
>/dev/dsk/c0t0d0s1     413639  250169  122107    68%    /var
>/dev/dsk/c0t1d0s7    8509324  339115 8085116     5%    /export/home
>/dev/dsk/c0t0d0s5    3007086  917816 2029129    32%    /opt
>/dev/dsk/c0t0d0s7    4031022  140916 3849796     4%    /usr/local
>swap                  643192     400  642792     1%    /tmp

This was the original setup on cyclone.  As you read through this email,
you will see that things have changed (for the better).

>Given the above, how do you suggest I deal with the matter of /var being
>set to 400 MB? Just looking at the allocation, I can reduce the storage
>allocated to /usr/local and transfer it to /var, or I can have the data
>stored on /export/home which you indicated was a less desirable solution.

Here is the implementation (i.e. what we did) of our recommendation:

  <login as ldm>
  ldmadmin stop
  <login as root>
  <move everything in /usr/local into /opt>
  cd /opt
  ufsdump 0f - /dev/dsk/c0t0d0s7 | ufsrestore rf -
  rm restoresymtable
  umount /usr/local
  newfs -i 8192 /dev/dsk/c0t0d0s7

  <move everthing in /export/home to what was /usr/local>
  mount /dev/dsk/c0t0d0s7 /mnt
  cd /mnt
  ufsdump 0f - /dev/dsk/c0t1d0s7 | ufsrestore rf -
  cd /mnt
  rm restoresymtable
  cd /
  umount /mnt

  cd /etc
  <edit vfstab>

  /dev/dsk/c0t1d0s7       /dev/rdsk/c0t1d0s7      /usr/local    ufs     2       
yes -


  /dev/dsk/c0t1d0s7       /dev/rdsk/c0t1d0s7      /opt          ufs     2       
yes -

  Do the same thing for the entry for the /export/home entry in vfstab:


  /dev/dsk/c0t1d0s7       /dev/rdsk/c0t1d0s7      /export/home    ufs     2 yes 


  /dev/dsk/c0t1d0s7       /dev/rdsk/c0t1d0s7      /data           ufs     2 yes 


  newfs -i 8192 /dev/dsk/c0t1d0s7

  After this, what was /export/home will be empty.

  cd /
  mkdir /data
  mount /data
  mkdir /data/ldm
  mkdir /data/ldm/mcidas
  chown ldm:ldmgroup /data/ldm /data/ldm/mcidas
  chmod 775 /data/ldm/mcidas
  cp -p /var/data/mcidas/* /data/ldm/mcidas
  cd /var
  rm -rf data

  cd /usr/local/ldm
  rm data
  ln -s /data/ldm data
  ln -s /data/ldm/logs logs

  <login as 'ldm'>
  <edit etc/pqact.conf>
  change all occurrances of /var/data/mcidas to data/mcidas
  ldmadmin mkqueue 
  touch data/logs/ldmd.log
  ldmadmin start

You see, piece of cake ;-)

This leaves you with the following disk configuration:

cyclone# df -k
Filesystem            kbytes    used   avail capacity  Mounted on
/proc                      0       0       0     0%    /proc
/dev/dsk/c0t0d0s0      96455   40767   46043    47%    /
/dev/dsk/c0t0d0s6     877790  665199  151146    82%    /usr
fd                         0       0       0     0%    /dev/fd
/dev/dsk/c0t0d0s1     413639   22750  349526     7%    /var
/dev/dsk/c0t0d0s5    3007086 1058723 1888222    36%    /opt
/dev/dsk/c0t0d0s7    4031022  340533 3650179     9%    /export/home
swap                  673024      16  673008     1%    /tmp
/dev/dsk/c0t1d0s7    8509324  180144 8244087     3%    /data

Now, you can put LOTS of data in /data/ldm...  User's home directories
will go into /export/home where there is plenty of room.  /usr/local
is now the same as /opt.

There are several things you don't need and several security holes that
we fixed for you.  This was done by commenting out entries in
/etc/inetd.conf as 'root':

name           (these are just annoying)
uucp           (these are just annoying)
discard        (these are just annoying)
daytime        (these are just annoying)
chargen        (these are just annoying)
100083/1       (tooltalk database    is a known security problem)
100221/1       (KCMS Profiler Server is a known security problem)
100068/2-5     (calendar manager     is a known security problem)

kill -HUP 148  (148 is the process ID of inetd)

>My current plan is to still have cyclone ingest and decode the data and
>then to have students using the linux machines to access and display the
>data products.

This sounds like a good, usable plan.

The next thing to do is to finish off the McIDAS installation:

<login as 'mcidas'>

cd mcidas7.6/update
ftp ftp.unidata.ucar.edu
  cd unix/760/bugfix
  get mcupdate.tar.Z
cd ../src
make all
make install.all

cd ~/data

<edit LOCDATA.BAT and change all occurrences of fully_qualified_hostname
to cyclone.natnet.du.edu>

<edit LOCAL.NAM and set the directories for McIDAS data files to

cd ~/workdata
<start a McIDAS-X session>


TE XCDDATA "/data/ldm/mcidas




The TE XCDDATA... set of commands sets up the McIDAS environment
for running of the XCD decoders.

The BATCH LSSERVE.BAT line sets up the ADDE data sets for the remote
ADDE server.

The ROUTE REL commands enable the creation of composite images by
way of ROUTE PostProcess BATCH invocations.  This works because I
configured ~ldm/decoders/batch.k yesterday.

<edit mcscour.sh and set MCHOME to /export/home/mcidas.  This sets up
McIDAS data file scouring>

setenv EDITOR vi
crontab -e
<add entry for scouring McIDAS data files>

<login as 'ldm'>
ldmadmin stop
cd etc
<edit ldmd.conf and add the entry:

exec   "xcd_run MONITOR"

<edit pqact.conf and uncomment out the entries for XCD stuff that I
added yesterday:

# Entries for XCD decoders
DDPLUS|IDS      ^.*     PIPE
        xcd_run DDS
HRS      ^.*    PIPE
        xcd_run HRS

Restart the LDM:

ldmadmin start

You now have XCD decoders producing McIDAS data files.  The data will
be scoured by virtue of the crontab entry above made as the user 'mcidas'.
Your remote ADDE server is working; I was able to point to it and list
and load data back to my machine here at Unidata.

>Thanks in advance.

You are welcome.  There may be a couple of issues related to proper
execution of some McIDAS commands.  I have discovered a bug in the
Sun SC5.0 Fortran compiler that affects several routines here at Unidata.
I see that the same problem is showing up on your system by getting:

PTLIST RTPTSRC/SFCHOURLY                                               
Row :      1  Col :      8                                             
Program terminated, segmentation violation 

The other programs that will probably have problems are GRDCOPY and GRDDISP.
I am working on finding a fix/workaround for the SC5.0 bug.

>P.S. NLDN data are now being ingested as well as the mcidas and FSL2

OK, I see that:

Pos      Description                        Schema  NRows NCols  Date
------   --------------------------------   ------  ----- ----- -------
     9   SAO/METAR data for   16 SEP 1999   ISFC       72  4500 1999259
    10   SAO/METAR data for   17 SEP 1999   ISFC       72  4500 1999260
    19   Mand. Level RAOB for 16 SEP 1999   IRAB        8  1300 1999259
    20   Mand. Level RAOB for 17 SEP 1999   IRAB        8  1300 1999260
    29   Sig.  Level RAOB for 16 SEP 1999   IRSG       16  6000 1999259
    30   Sig.  Level RAOB for 17 SEP 1999   IRSG       16  6000 1999260
    39   SHIP/BUOY data for   16 SEP 1999   ISHP       24  2000 1999259
    40   SHIP/BUOY data for   17 SEP 1999   ISHP       24  2000 1999260
    59   SYNOPTIC data for    16 SEP 1999   SYN         8  6000 1999259
    60   SYNOPTIC data for    17 SEP 1999   SYN         8  6000 1999260
    69   PIREP/AIREP data for 16 SEP 1999   PIRP       24  1500 1999259
    70   PIREP/AIREP data for 17 SEP 1999   PIRP       24  1500 1999260
    79   NLDN data for        16 Sep 1999   NLDN     1000  1000 1999259
    80   NLDN data for        17 Sep 1999   NLDN     1000  1000 1999260
    89   Hourly Prof data for 16 Sep 1999   WPRO       48    35 1999259
    90   Hourly Prof data for 17 Sep 1999   WPRO       48    35 1999260
    99   6-min Prof data for  16 Sep 1999   WPR6      480    35 1999259
   100   6-min Prof data for  17 Sep 1999   WPR6      480    35 1999260

I will look at your setup more tomorrow...


NOTE: All email exchanges with Unidata User Support are recorded in the Unidata inquiry tracking system and then made publicly available through the web. If you do not want to have your interactions made available in this way, you must let us know in each email you send to us.