Re: Giving up on x buffers - losing files

From: Jon Noack <noackjr_at_alumni.rice.edu>
Date: Sun, 27 Jun 2004 13:17:22 -0500
On 06/26/04 23:38, Don Lewis wrote:
> On 26 Jun, Arjan van Leeuwen wrote:
>> Sometimes, particularly after doing a lot of file writes (i.e.
>> compiling a lot of ports, building world and mergemastering, etc),
>> I get the 'Giving up on x buffers' message on shutdown, and my
>> filesystems come up dirty when I restart.
>> 
>> <snip>
>> 
>> So, why does this happen? And how do I prevent it from happening?
>> This definitely does _not_ sound like something I want my servers
>> to do when 5.x goes -STABLE.
> 
> I've mentioned this a couple of times on this list in the last six 
> months or so.  The last time was in the last couple of weeks.  I can 
> reliably trigger this problem with mergemaster.
> 
> I'm pretty sure that the problem relates to soft updates and how the 
> file system syncer is shut down, which leaves unresolved dependencies
>  that keep a number of dirty blocks from being flushed to disk at the
> end of the system shutdown.
> 
> I have some ideas on how to fix the problem, but I haven't had the
> time to work on it and nobody else has stepped up with a fix.
> 
> I am able to reliably work around the problem by running the sync 
> command and waiting a short while after running mergemaster and
> before shutting down or rebooting the machine.

If you're running X be sure to end your session instead of restarting 
directly.  I usually end the session, switch to a console, manually sync 
as Don described, wait 30 seconds, manually sync again (just to be 
sure), and then shutdown.  Actually, I wrote a script to do it (sync; 
sleep 30; sync; shutdown -r now).

I only had the problem a couple times, but never since starting to 
manually sync.

On a more humorous note, trying to save time by restarting both your NFS 
server and client workstation at the same time is a really bad idea, 
especially when your home directory is NFS mounted... ;-)

Jon
Received on Sun Jun 27 2004 - 16:17:32 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:59 UTC