On the ext3-users list, NFS maintainer Neil Brown mentions about an ext3 bug where time stamps of some blocks are wrong so old data doesn't get flushed out by bdflush when it should, this seems to increase memory pressure on busy fileservers and database servers and causes writestorms when the data is forced out on a full journal I have filed a bugzilla report (against 7.3) for this, though even the RH 9 kernel doesn't have this fix https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=90191 I see it in rawhide but given the recent introduction of exec-sheild in rawhide, it's not appropiate to use this in production. Arjan/Matt/Bill, any chance of an errata kernel which incorporates the bdflush fix Regards, Yusuf -- Yusuf Goolamabbas yusufg@xxxxxxxxxxxx
