Boot time take too long - Veritas Volume Manager

This is a discussion on Boot time take too long - Veritas Volume Manager ; SEVM 2.6 and VxFs 3.3.2 on a UE10000 SUN server takes more than 1.5 hours to boot. SEVM takes more than 35 minutes to scan the EMC disks and build the disk groups. Then, VxFs does FSCK in serial. We ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: Boot time take too long

  1. Boot time take too long


    SEVM 2.6 and VxFs 3.3.2 on a UE10000 SUN server takes more than
    1.5 hours to boot. SEVM takes more than 35 minutes to scan the EMC
    disks and build the disk groups. Then, VxFs does FSCK in serial. We
    have more than 800 file systems in about 8TB of EMC SYMTRIX.

    Is there a way to improve boot time of this server. Thanks....

  2. Re: Boot time take too long


    "Zaki AL-Husain" wrote:
    >
    >SEVM 2.6 and VxFs 3.3.2 on a UE10000 SUN server takes more than
    >1.5 hours to boot. SEVM takes more than 35 minutes to scan the EMC
    >disks and build the disk groups. Then, VxFs does FSCK in serial. We
    >have more than 800 file systems in about 8TB of EMC SYMTRIX.
    >
    >Is there a way to improve boot time of this server. Thanks....


    Well - well -well, Let's see. Over 800 file systems and
    8TB of data being done in 1 1/2 hr. is the equivalent of
    0.11 second average per fs. and I'm not mentioning checking
    on all other peripherals you might have attach to this
    E10000. It will be good to see other people opinion
    running E10000 to see how your machine boot time compare
    to their.
    Vinnie German.


  3. Re: Boot time take too long


    "Vinnie German" wrote:
    >
    >"Zaki AL-Husain" wrote:
    >>
    >>SEVM 2.6 and VxFs 3.3.2 on a UE10000 SUN server takes more than
    >>1.5 hours to boot. SEVM takes more than 35 minutes to scan the EMC
    >>disks and build the disk groups. Then, VxFs does FSCK in serial. We
    >>have more than 800 file systems in about 8TB of EMC SYMTRIX.
    >>
    >>Is there a way to improve boot time of this server. Thanks....

    >
    >Well - well -well, Let's see. Over 800 file systems and
    >8TB of data being done in 1 1/2 hr. is the equivalent of
    >0.11 second average per fs. and I'm not mentioning checking
    >on all other peripherals you might have attach to this
    >E10000. It will be good to see other people opinion
    >running E10000 to see how your machine boot time compare
    >to their.
    > Vinnie German.
    >

    Do you have extraneous sd.conf entries?
    Your EMC rep should be able to provide you with the
    correct revs of your JNI driver, vxvm rev, EMC firmware, Solaris
    patch rev, etc. They really do need to all 'match' to get
    good performance.
    You spent two million dollars and can't afford the real version
    of VXVM? It costs less than the sales tax on the server.
    Hey, it could be worse, a lot of mainframes take more than a
    day to reboot.
    Have you looked at your diag stuff in the EEPROM? Turn it all off.
    Check for network problems on the console. boot net isn't
    fast to begin with, but a bad ethernet connection ( do a netstat
    -in ) will really slow you down.
    How long does it take to do a /usr/symmapps/bin/inq -et? This will
    tell you how much time the fca driver spends finding drives.
    I you've renumbered your controllers or LUNS or anything,
    Solaris and VX could be waiting on many drives to respond to
    inq's that they'll have to time out on. One way to sort that out
    is to do a 'grep "not present" /var/adm/messages | wc -l' and
    see how many disks he looks for that never show up.
    vxdisk list | grep error | wc -l
    will do almost the same thing.
    Good luck.




  4. Re: Boot time take too long


    Thank you al for your response. I finally got a grip of the situation. I have
    managed to reduce
    The boot time form 70minutes to 27 minutes, and this is how I did it

    Boot to single user mode used to take about 35 minutes. I removed all unneeded
    entries form the sd.conf file and I used the parent option instead of the
    SCSI option. That saved me 15 minutes. I then removed the DMP directories
    which save me another 10 minutes. Now, to save time going to multiuser mode
    I had to
    Intercept /sbin/mountall and call another script to do parallel FSCK on VXFS
    file systems. That saved me about 20 minutes. I have one more challenge
    which is if I issue /etc/halt the machine takes bout 30 minutes before I
    see the OK> prompt. I know that it is spending the time unmounting the file
    systems


+ Reply to Thread