mschilli at vss
Feb 19, 2003, 12:51 PM
Post #3 of 3
(516 views)
Permalink
|
The particular case that I am working with, is that even fuser won't
kill kernel processes (exported NFS mountpoints)..
Matt Schillinger
mschilli[at]vss.fsi.com
On Wed, 2003-02-19 at 10:58, Alan Robertson wrote:
> Matt Schillinger wrote:
> > My understanding is that heartbeat does not expressly manage mounts, but
> > the Filesystem start/stop script manages them. If there was a heartbeat
> > stop (graceful), but something hung up and the unmount was unable to
> > complete, would heartbeat know, to prevent the mounting from the
> > secondary system? Would Filesystem unmount returning '1' cause the
> > process to stop and the secondary not to mount the system? or would that
> > cause a stonith?
>
> I'll go read the code...
>
> The filesystem script will do a kill on any processes which have the
> filesystem open.
>
> But, it won't persist and kill the stubborn ones with -9 (a bug IMHO).
>
> I suspect heartbeat doesn't handle the return codes from resource scripts
> correctly.
>
> Here's what I think should happen here:
>
> 1) The Filesystem script should be more persistent
> 2) Heartbeat should check the return codes
>
> And perhaps...
>
> 3) Heartbeat should audit the resources to make sure they're
> released before declaring them to be released.
>
> There are three circumstances under which it releases resources:
>
> the ip-request script
>
> shutdown
>
> hb_standby request
>
> I believe that all of these need to be cleaned up.
>
>
>
> --
> Alan Robertson <alanr[at]unix.sh>
>
> "Openness is the foundation and preservative of friendship.... Let me claim
> from you at all times your undisguised opinions." - William Wilberforce
--
Matt Schillinger
System Administrator
FlightSafety International
mschilli[at]vss.fsi.com
314-551-8403
|
