HP-UX 11.31 how do you mount a file system in an ordinary file?

[Hilary Palmer]

Hello,

I am trying to create a file system in an ordinary file.

• mkfs_vxfs has the option using "-o X"
• mkfs_hfs has the option using "-d"

But when I go to mount the file, mount tells me that my file "is an invalid operand".

What am I doing wrong?  The mkfs command require the file to exist, so do I need to make this file in some special way?  I have tried doing a simple touch command, and I also tried creating the file with dd like I would in Linux.

Thanks,

Hil

[Ken Grabowski]

Before somebody can tell you what you may be doing wrong, you need to give the detail about what your doing. Show the actual command lines you used and any response from the system.  Check /var/adm/syslog/syslog.log for any corresponding error messages.

Making a file system inside a file instead of on a storage device located at a hardware special device file location is for debugging purposes only.  What are you trying to create a file system inside of a file for?  Are you possibly trying to do something along the lines of creating a mountable ISO image? If so, you can find the instructions  in the ignite users guide.

[Hilary Palmer]

# touch hilfs
# mkfs -F hfs -d hilfs 1024000
mkfs (hfs): hilfs - 1024000 sectors in 2000 cylinders of 16 tracks, 32 sectors
1048.6Mb in 125 cyl groups (16 c/g, 8.39Mb/g, 1280 i/g)
Super block backups (for fsck -b) at:
16, 8240, 16464, 24688, 32912, 41136, 49360, 57584, 65808, 74032,
82256, 90480, 98704, 106928, 115152, 123376, 131088, 139312, 147536, 155760,
163984, 172208, 180432, 188656, 196880, 205104, 213328, 221552, 229776, 238000,
246224, 254448, 262160, 270384, 278608, 286832, 295056, 303280, 311504, 319728,
327952, 336176, 344400, 352624, 360848, 369072, 377296, 385520, 393232, 401456,
409680, 417904, 426128, 434352, 442576, 450800, 459024, 467248, 475472, 483696,
491920, 500144, 508368, 516592, 524304, 532528, 540752, 548976, 557200, 565424,
573648, 581872, 590096, 598320, 606544, 614768, 622992, 631216, 639440, 647664,
655376, 663600, 671824, 680048, 688272, 696496, 704720, 712944, 721168, 729392,
737616, 745840, 754064, 762288, 770512, 778736, 786448, 794672, 802896, 811120,
819344, 827568, 835792, 844016, 852240, 860464, 868688, 876912, 885136, 893360,
901584, 909808, 917520, 925744, 933968, 942192, 950416, 958640, 966864, 975088,
983312, 991536, 999760, 1007984, 1016208
# mount /root/hilfs /mnt
mount: /root/hilfs is an invalid operand

The problem that I am facing is that I am running a backup off of a Compellent SAN.  In one of the filesystems that I am backing up I have 1000's of tiny files.  Using fbackup I am seeing backups vary between 45 mins (what I used to see off the EVA) and 3 hours.  So I was hoping to be able to create a filesystem in a file.  When the backup runs I will umount the filesystem and backup that single file.  Backing up the databases which are large files are consistently backing up within 30 minutes which I expect.

[Ken Grabowski]

Interesting concept. Haven't seen anybody try this before. Not sure if it will work, or speed up your backup. If the SAN underneath your storage is slow,  making a single big file to replace the file system is probably not going to speed it up.

Did you read mount_hfs?  I would suggest at least adding -F  hfs after the mount and before your special device.

[Hilary Palmer]

I would assume it would since my backups with large files are consistent.  The ton of little files are killing me!  Just trying to think outside of the box, since Dell is not being very helpful.  They are not seeing anything that is "off", and I am burning SSD's with this filesystem.

My backup for the database is 425GB and takes about 27 mins.  The backup with the little files is 186GB in size.

I did try adding -F hfs.  But I get the same message.

[Ken Grabowski]

What are you backing up too with fbackup?  What is in your fbackup configuration file?

[Hilary Palmer]

MSL4048 G3 with 2 FC LTO5 drives.

fbackup.config...

blocksperrecord 4096
records 64
checkpointfreq 4096
readerprocesses 6
maxretries 5
retrylimit 5000000
maxvoluses 200
filesperfsm 2000

[Ken Grabowski]

Try:

blocksperrecord 256
records 32
checkpointfreq 1024
readerprocesses 6
maxretries 5
retrylimit 5000000
maxvoluses 200
filesperfsm 2000

Make sure your file system is quiesced. Open files could be part of your issue.  

[Bill Hassell]

fbackup is dramatically affected by the config file settings.

This should provide a fairly big jump in performance:

blocksperrecord 4096
records 64
checkpointfreq 8192
readerprocesses 6
maxretries 5
retrylimit 5000000
maxvoluses 200
filesperfsm 5000

The fixes for massive file counts is to create large records (4096), don't run checkpoint too often (8192) and don't create setmarks very often (5000).

I think what you will find to much simpler is to just run tar to collect all the files and directories into one giant file. Then backup that one tar file.

[Hilary Palmer]

Yeah...  I considered the tar/gzip option...  It took 2 hours to do the backup of the problem filesystem, and I would need to steal more space from the array for a perminant landing zone since it was 53GB.

[Ken Grabowski]

Pretty sure your problem comes down to tweaking your fbackup settings and quiesceing the file system.  If you have active files fbackup can stall and retry until it hits its retry limit. An active open files may be skipped.

[Hilary Palmer]

Well...  I was able to get the tape drive to flag an error code.  Drive is faulty.  Gettting a replacement.

Thanks for the assistance.

But back to the subject...  Anybody know how to do that?