Coda File System

Re: max venus cache size?

From: Jan Harkes <>
Date: Mon, 13 Sep 2004 08:58:57 -0400
On Fri, Sep 10, 2004 at 03:35:30AM -0500, Brian Finney wrote:
> It seems that sence my average file is going to be so much larger than
> what has been assumed I will also be able to serve a much larger
> amount of data than is generally referenced in most of the docs and
> during the install scripts for the givin rvm sizes.  Is this true?


> Also is there any way that I would be able to check how much of the
> rvm is being used so that I can better tweek these sizes in the
> future.

There are no single easy way. There is a volutil rvmsize command which
can be used to get rvm usage numbers for specific volumes. Once in a
while (every N thousand operations) the server logs to /vice/srv/SrvLog
'RDS Statistics', and shows total the current bytes allocated/free. This
periodic statistics dump can also be triggered with volutil printstats.

The client should also have something similar, I think it was 'vutil
stats' and the statistics should show up in /usr/coda/etc/venus.log
(or /var/log/coda/venus.log)

> also during the install the install script mentioned that for
> primarily read only files systems it may actually be better to use
> files for the log and rvm, is this still considerd true?

Yes, files do seem to be better, strangely enough using raw partitions
'should' reduce double caching in memory and make fsync faster. However,
fsync is about as fast either way and with a file backend we can use
mmap and we don't end up writing unmodified pages to swap, which makes
the server a lot more when there is some memory pressure.

Received on 2004-09-13 09:02:21