I had posted some questions about advfs in an earlier post and thought I would
update everyone with what I've found out so far. Also, I have some more
questions.
**********************MY EARLIER QUESTIONS:
>I was at the Fall DECUS and one of the DEC people told me that there was
>a internal DEC cookbook on tuning ADVFS.
there apparently is an internal DEC doc called "A Hitchiker's Guide to
ADVFS" that is on some internal DEC WWW site. We have not been able
to verify its existence by getting a copy of it yet. sorta like
searching for the holy grail.......
>Do the VFS tuning and configuration considerations described in
>"System Tuning and Performance Management" (AA-Q0R3D-TE), section 3.6.1.1
> apply when using ADVFS? Do the UBC Subsystem considerations apply when using
>ADVFS?(section 3.4.1) Do the Metadata and Namei Buffer Cache considerations
>apply?(section 3.4.1.3)
our DEC support says yes to these questions.
>The Release Notes recommend extending the metadata extent size (using the
>'addvol -x #pages device domain_name' command), but do not tell how to display
>the current size. How do we display the current size of the metadata table for
>a particular file domain?
To find out the current size, cd to the filesystems directory, then do
the following...
cd .tags
showfile -x M-6
Look at the pageCnt column...ignore the first row. If it says 1024,then
you are ok. If it says something smaller, such as 256, then you need
to backup, rmfset, rmfdmn, mkfdmn, mkfset, restore.
-thanks to Rich Jackson for this info.
*********************************************
NEW QUESTIONS:
1) What commands can we use to tell how well AdvFS file I/O caching
is working. (i.e., we'd like to see %hit rates, etc. for AdvFS)
2) dxadvise says our "overall soft page fault rate" is too high.
How can we determine how much of the page fault activity is coming
from file I/O vs. program memory use?
3)dxadvise showed that our wired memory had grown to 116M. In three days,our
wired memory has grown from 35M to 85M. vmstat currently shows
MIPC = 52M of memory. The usage appears to be out of bucket #4(element size
128). Our system has 192M physical memory. The system grinds to a halt shortly
after this happens. What are some areas that could cause wired memory to grow so much/so fast?
TIA,
Randy Marchany
VA Tech Computing Center
Blacksburg, VA 24060 USA
INTERNET: randy.marchany_at_vt.edu
Received on Mon Jan 15 1996 - 04:34:32 NZDT