(Illustration by Gaich Muramatsu)
On Tue, Jul 08, 2003 at 07:17:05PM +0200, Peter Sch?ller wrote: > > There are volume wide RW locks. When a client is connected over a modem > > we use trickle reintegration which doesn't (shouldn't) lock down the > > volume as the file is sent to the server in small fragments and the > > actual 'commit' is only performed once the complete file is already > > present on the server. > > Ah. Is there a way tro control/monitor this behavior? (As in this test > case it was hogging the server for an extended period of time.) Not that I know of. I prefer to work with large client caches, so we only need to fetch updated data during a hoard walk. And with writeback caching (write-disconnected) so we push updates to the server in the background. I guess this makes me not really notice when a volume happens to have an exclusive write-lock. All writes happen in the background and most changes are picked up by the hoard daemon. JanReceived on 2003-07-08 12:50:53