Hello, I have a unique deployment scenario where the marriage of ZFS zvol and UFS seem like a perfect match. Here are the list of feature requirements for my use case:
* snapshots * rollback * copy-on-write * ZFS level redundancy (mirroring, raidz, ...) * compression * filesystem cache control (control what's in and out) * priming the filesystem cache (dd if=file of=/dev/null) * control the upper boundary of RAM consumed by the filesystem. This helps me to avoid contention between the filesystem cache and my application. Before zfs came along, I could achieve all but rollback, copy-on-write and compression through UFS+some volume manager. I would like to use ZFS but with ZFS I cannot prime the cache and I don't have the ability to control what is in the cache (e.g. like with the directio UFS option). If I create a ZFS zvol and format it as a UFS filesystem, it seems like I get the best of both worlds. Can anyone poke holes in this strategy? I think the biggest possible risk factor is if the ZFS zvol still uses the arc cache. If this is the case, I may be double-dipping on the filesystem cache. e.g. The UFS filesystem uses some RAM and ZFS zvol uses some RAM for filesystem cache. Is this a true statement or does the zvol use a minimal amount of system RAM? Lastly, if I were to try this scenario, does anyone know how to monitor the RAM consumed by the zvol and UFS? e.g. Is there a dtrace script for monitoring ZFS or UFS memory consuption? Thanks in advance, Brad _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
