fcdfdata012 ( dual 1.4 GHz, 1.8 Tbytes disk, Gbit ethernet )
   is set up with a stager account for holding the rman backups.
   It received daily copies of the rman files from b0dau35.    
   The copy, checksum and gzipping of daily rman's is now automatic.

STKEN Enstore tapes are being written to dual tapes on days 1, 11, 21 monthly.
    So far we have written tapes for Feb 1, 11, 21, Mar 1.

Immediate plan :

    Continue to run the fcdfdata012 archives in parallel with amanda,
        to test reliability, and for redundancy.
        fcdfdata012 can hold about 3 weeks' of rmans comfortably.

    Continue to write full rman's to tape regularly
        ( 01, 11, 21 of each month to dual tapes )

Specifics :

    See
        http://fcdfhome.fnal.gov/usr/kreymer/dbback/README.html

    Ugly details are in /cdf/home/www/usr/kreymer/dbback/LOG

    The pushrman script runs on b0dau35 daily, at 16:00, done in 8 hours.
        ( I would start it at 04:20, but amanda starts at 06:00.)

        Gets kerberos ticket
        Checks /ora1/admin/cdfonprd/bdump for today's log file
            Waits for successful rman completion if necessary
        Copies rman files to fcdfdata012 dated directory with rcp
            Takes about 4 hours elapsed time, CPU load is about 1/2 processor
        Spawns remote checksum/gzip job in the background
            Checksum runs about 1.5 hours, gzip about 4.5 hours on '012 .
        Runs local checksum
            About 4 hours, CPU load is about 1/2 processor (top shows 12%)
        Copies local copy log and checksum log to fcdfdata012

    The rman files are about 140 GBytes raw, and 52 GBytes compressed.
    I use gzip -1 , this produces 15% larger files than the default -6,
        at half the CPU cost.
        gzip -9 (or -best) provides no further reduction, at 8 times the cost,
        or 16 times the cost of gzip -1 .
        See  HOWTO.gzip
        amanda jobs have recently been modified from -9 to -1

Planning :

    The rman directory size has increased by 1.8 GB the last 12 days,
    or about 55 Gbytes per year if blindly extrapolated.

    Tapes needed for two copies every 10 days is about 20 tapes/yr.
        Tape capacity is about 200 GBytes.

    fcdfdata012 reliability - OK so far.
        If we're worried, swap it for another DCache file server with
        a proven track record.

    Isolation contengency - What if we have to cut the B0/FCC network link ?

        We still have the daily rman's on the local disk.

        We could set up an additional file server in B0
            ( no need for 2 TB disk )

        We could put an external 200 GByte USB-2 disk on fcdfdata012,
           holding recent rman's and archives. In a disaster, carry it to B0.
           These are inexpensive, about $300 retail.