Re: Overlong wait for old active object to go away

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Fri, 2 Sep 2011, David Howells wrote:

> Mark Hills <mark@xxxxxxxxxxx> wrote:
> 
> > In dmesg "Overlong wait for old active object to go away" (see full 
> > message below), and an application will hang or lock, eg. blocking on 
> > read().
> > 
> > Sometimes it unblocks, sometimes it seems to hang indefinitely. Is 
> > there a higher level explanation for what is going on here?
> 
> I suspect that the OBJECT_RECYCLING state doesn't complete the 
> transition to the OBJECT_DYING state since we saw that we had to wait 
> for outstanding operations to complete at the time OBJECT_DYING was 
> processed.  I think there's probably a race in waking up the object 
> state machine when the last operation completes.
> 
> It would be useful you could try is grepping in /proc/fs/fscache/objects 
> for the objects named in the error text.  The object debug ID is 
> right-aligned in the first column without the "OBJ" attached.

Sure, I'll try this next week and hopefully get you more info. I was 
hoping for some useful debug information like this, thanks.
 
> I think I see a gap in my testing:-(.  I've mainly tested against a 
> static set of files on the server for reproducibility - but that doesn't 
> test these paths.

This could reasonably explain why such problems have gone unnoticed. I'm 
trying to use cachefs for day-to-day work; reads and writes.

My /home is on NFS, which probably explains why I get hit hard. Perhaps 
give that a try for a bit if you are keen to try and see these problems.

> I have manually done individual file alterations, but that doesn't 
> stress things.

I see very common lockups on Opera's (web browser) cache directory. I 
suppose the nature of a cache directory is that it is 'stressful' on reads 
and writes on small files. I also see it on program source code files when 
I am developing.

In these cases they're being modified only by the one client, which is 
also the client with cachefilesd running.

Thanks for the useful info.

-- 
Mark

--
Linux-cachefs mailing list
Linux-cachefs@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cachefs


[Index of Archives]     [LARTC]     [Bugtraq]     [Yosemite Forum]
  Powered by Linux