My pseudo worm is getting crouded ...

My pseudo worm is getting crouded ...

[Arie]

Hi,

My main file system resides on a cached pseudo worm device which 
currently has a 1GB disk for the worm part.  I suspect it may be 
getting full.

I'd like (very much) to add a new disk to the SCSI chain, go
into config mode, redefine `filesys main' and be merry.

Can this be done?  Did anyone do it and survive?

My source browsing brought me to the (maybe wrong) conclusion that
redefining the filesys in config mode seems to be feasible.
config() will add the drive to the list of `main' and devinit()
it, which will write an empty bitmap at the end of the drive.

The question is: will the fake worm gracefully overflow into the
newly added drive?

Any hints will be mightily appreciated!  Loosing my `main' ain't
a happy prospect.

thanks

blumzi

  -~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~
        ___ ___ ___             Arie Blumenzweig
    __ /__ /  //_               nSOF Parallel Software Ltd.
   / /___//__//                 48800 Givat Hashlosha, Israel
                                Tel: +972 (3) 938-8551
   blumzi@nsof.co.il            Fax: +972 (3) 938-8552
  -~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~

My pseudo worm is getting crouded ...

[beto]

In <9603251139.AA10775@nsof.co.il>
	blumzi@nsof.co.il (Arie Blumenzweig) wrote:

> Hi,
> 
> Can this be done?  Did anyone do it and survive?
> 

I did something similar but not exactly the same. 
I was using a MIPS machine as a file server with a 
4G SCSI disk as the main file system. I wasn't using
a fake worm partition becuase the data there was
also inmutable. Eventually I ran out of space, so I was
able to attach another SCSI disk then redefine main as
the concatenation of the old and a new 4G disk. 

The procedure was:

0) recompile the kernel with the new file system
    size. the kernel does not calcualte the size
    of the file system based on main definition, it 
    read it from the super block.
	
1) config mode 
	- filsys main (wOLDwNEW)
2) normal mode
	- then run check free which will recreate the
	   free list and hopefully add the new blocks to the
	   free list.

However, what you want do it's a bit different:

	- You won't have to run check free becuase 	
	   bufalloc will add more block to the free
	   list of demand, which is good.
	- If had a look to the fworm implementation and
	   it uses a part of the device to detect if a block
	   is written twice or read without been written first.
	   So you would probably have to ream both disk
	   again, which would destroy your data :-(.

It looks to me that the easiest thing is to backup
your files restart a new file system with the
two disks, restore the backup and wait until
5 pm, however you'll loose the history. Another
posibility is to add a command to the kernel to
combine both disks into a new fworm, but please
backup your disk first.