Subject: | |
From: | |
Reply To: | Paveza, Gary |
Date: | Fri, 14 Jul 2000 09:26:06 -0400 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Jeff,
One of the things a previous admin here has done is to preallocate space on
LDEV 1 which is reserved. We do this in one extent as well (saves us the
trouble of trying to find that one contiguous chunk when doing an upgrade).
While this doesn't directly resolve the issue you mentioned, it is one way
to set aside space for those emergencies. You can simply build the files on
each LDEV you want to reserve on.
:listf dummy.pub.sys,2
ACCOUNT= SYS GROUP= PUB
FILENAME CODE ------------LOGICAL RECORD----------- ----SPACE----
SIZE TYP EOF LIMIT R/B SECTORS #X MX
DUMMY 128W FB 0 130000 1 130000 1 32
-------------------------------------------------------------
Gary L. Paveza, Jr.
Technical Services Manager
(302) 761-3173 - voice
(800) 217-5808 - pager
-----Original Message-----
From: Jeff Kell [SMTP:[log in to unmask]]
Sent: Thursday, July 13, 2000 10:19 PM
To: [log in to unmask]
Subject: Re: [HP3000-L] How can Transient free space be not
full on a disk ifperm. freespace is?
Jean Huot wrote:
>
> I have a defective system disk which I filled up with garbage.
> When I run disfree.pub.sys, it tells me I have 533 Mbytes free for
> transient free space and 0 Mbytes for permanent free space. How
> can that be possilble?
When you define a volume, you give it a maximum percentage of space
that
may be allocated to permanent and transient space (as others
have pointed out). I think this point got across, but I have an old
gripe about this I'd like to drag out of the closet.
VOLUTIL will happily let you allocate 100% to each type allocation,
and
has no problems with you making limits with a sum > 100%. However,
VOLUTIL will *NOT* let you make limits with a sum < 100%.
Personally, I'd like to have some "reserve" space that could be
allocated just after you experience free space problems but before
you are forced to reload, reorganize, or order new spindles. This
is
particularly nasty on nearly-full, fragmented spindles. There is no
way
to tell MPE to "please use another drive as I'm nearly full" as the
end
result is that it will continue to allocate space as long as space
is
available. This is *especially* nasty on LDEV 1 where the initial
VSM
transient allocations are made (before mounting members),
and the xxxxxx sectors of contiguous space required by UPDATE, and
the now-defunct (I think) bias of allocations away from LDEV 1 as it
approached 50% total allocation. Which refreshes my memory -- under
the
old 50% rule it forced transient away from LDEV 1, and if you only
had a
couple of spindles in the system volume set, the members would fill
up
while ldev 1 sat there with loads of free space. This was the root
of
the 'fork() of death' problem (search the archives).
Why does VOLUTIL force you to allocate the whole disc? Having some
space in reserve is the difference in a show-stopping reload or
other
drastic measure versus a volutil reallocation.
Jeff Kell <[log in to unmask]>
|
|
|