Date: Fri, 16 Dec 2005 14:37:26 +0100 From: Michael Schuh <michael.schuh@gmail.com> To: Eric Anderson <anderson@centtech.com> Cc: Ivan Voras <ivoras@fer.hr>, freebsd-geom@freebsd.org Subject: Re: Questions about geom-gate and RAID1/10 and CARP Message-ID: <1dbad3150512160537y4a944f81u@mail.gmail.com> In-Reply-To: <43A18CAB.6020705@centtech.com> References: <1dbad3150512090509l2ab08e03k@mail.gmail.com> <43998C95.3050505@fer.hr> <1dbad3150512130421i5278d693g@mail.gmail.com> <439ED00E.7050701@centtech.com> <1dbad3150512150355r576b1b0j@mail.gmail.com> <43A18CAB.6020705@centtech.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> There are others, like lustre, gfs, polyserve, etc, however none of them > work in FreeBSD at this point. A few people (including myself) have > started a project to port gfs to FreeBSD (gfs4fbsd project on > sourceforge). Oh i think this was a very good idea......i'm will watch this project. another good thing was gpfs from IBM....... > > > > I'm wondering actually if you couldn't actually do this with NFS, and > hacking some pieces together. Haven't thought through it, but seems > like maybe you could make the active writer an nfs server, that also > mounts it's own nfs share rw, but the nfs sharing would be on a virtual > interface, or at least the one that 'moves' with your failover. The > other machines would mount that nfs server's export ro, and when it > fails over, the one taking over would have to run a script to begin > serving that export rw to all, and it's own client would continue its > connection but now on it's new virtual interface. You'd also have to > have the ggate stuff set up, so that it was mirroring the original > 'master' disk, but when the failover occurred, you would quickly mount > your local mirrored disk rw, ignoring the 'unclean' message, begin a > background fsck, then start the nfs server on that mount point. You > would probably also have to fail the original drive in the mirror to > effectively 'fence' that node from making disk changes at the same time > the new master did. Oh yes, this was also possible, but i would slunk around the script/cron-he= ll. It wasn't really a hell, but a serious error source..... my prior idea was to mount union two different mounted nfs-servers..... like your suggestion, but here comes the next problem......unionfs.... and later the sync....... I hope a good solution for real HA is around the corner.... michael
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1dbad3150512160537y4a944f81u>