Coda File System

Re: write-back and new list of issues :)

From: Shafeeq Sinnamohideen <shafeeq_at_cs.cmu.edu>
Date: Thu, 14 Jun 2001 11:18:09 -0400 (EDT)
On Wed, 13 Jun 2001 ctest_at_neural.dlsemc.com wrote:

> Has write-back caching been finished?  this keeps happeing when i try to enable it:
> 
> i do-> cfs wbstart /coda
> and get->
> VIOC_BEGINWB: Resource temporarily unavailable
>   return code : -1

The server needs to allow write-back caching as well. 
This is done per volume by volutil enablewb/disablewb. 
Write-back support can be disabled for the entire server regardless of
volutil by starting it with the -nowriteback flag. 

Write-Back mode is only useful in the situation where one volume is only
accessed by one user, but from different workstations at different times. 

cfs wbstart gets you a write-back permit which is revoked when some other
client wants te read that volume. You have to get another permit to
re-enter write-back mode by doing this again.

cfs wbauto will periodically try to get a permit if you do not have one.

Write-Back mode is basically write-disconnected mode, with reintegrations
forced when the permit is revoked. The trouble with it is the definiton of
a reader is strict, so eg: doing ls -al /coda/usr/ will cause every user
volume in write-back mode to reintegrate. If you can live with less
consistency, write-disconnected mode is what you want.


> And (drumroll)... here's my list of new issues.
> 
> 
> issue #1
> if i do a  -> cfs listvol /coda
> i get -> operation not permitted
> if i do a 2nd  -> cfs listvol /coda
> i get the correct output (saying "connected" and stuff)
> 
> any ideas what causes this?  i can use my bash history easy, but i'm worried that it might be a symptom of a bigger problem, that i don't know about.

What system ? Does this happen only right after venus is started, or
always ?

Shafeeq
Received on 2001-06-14 11:18:41