On Thu, Jul 17, 2003 at 06:44:58AM -0700, Tim Hasson wrote:
> I created a 1GB RVM earlier and had to play around with rdsinit to fit it by 
> reducing the heapsize. However, if I turn mapprivate to 1 in server.conf file 
> and restart the server, tailing /vice/srv/SrvLog shows the error: 
> RVM_ENOT_MAPPED
> What do I have to do in this case to get RVM mapped?

Not sure, I know that linux only allows mmap from files and not raw
partitions.

> Also, is it possible to make/use more than a gig of RVM? Is it possible to 
> store like 100 gigs in 1 coda volume (supposing the average file size is 5-
> 10MB)?

In that case, a gig is probably already overkill. Most of the numbers
that we use are based on an average filesize of 16KB. As your files are
on average about 500x that size, you should be able to store a lot more
data. It is the per-file metadata overhead that is the problem for RVM,
http://www.coda.cs.cmu.edu/misc/rvm-usage.html.

If 50k files and 5k directories use about 40MB of RVM, then 1GB should
be able to hold the metadata for 1.2 million files and 125K directories.

But when I divide 100GB by 5MB per file, I get at most 20480 files and
if the average filesize is higher, the actual number of files becomes
even lower. So according to the previous calculations you should be fine
with as little as 20MB of RVM, but setting it up to use about 200MB RVM
data will give you more than enough room to grow.

Jan