Hello everyone, I'm building a highly available storage cluster, providing NFS service to several clients for virtualization storage, specifically vm VHD storage. I'm working with two Dell R300s (64 bit 4 core xeon, 8 gb ram), each with an LSI iSAS HBA, and three promise vtrak 310s storage chassis. The promise vtrak chassis have isas controllers that allow two servers to share storage. My plan is to create a ha zpool from two promise chassis, and enable replication to a third. My primary question is, is ha-zfs ready for production use or should I look into using an actual shared filesystem approach. I am interested in zfs only because it leaves open the option to expand the storage capacity and performance by simply adding a jbod. I understand that failing over the filesystem is a manual process, but I am ok with this.
Also, given the filesize I don't think there would be much benefit from throwing some SSDs into the pool as cache or log devices, but also wanted to ask if anyone thought there would be much benefit from this. I didn't copy this intentionally, but this is exactly what I'm going for: http://www.dataonstorage.com/dataon-solutions/107-zfs-ha-clustering-shared-storage-solution.html Also I'm just cutting my teeth on OpenSolaris, while I'm very familiar with FreeBSD and Linux any outside references you can point at are always appreciated. Thanks everyone! -Alex
