From: Dominik L. B. <do...@vb...> - 2005-04-22 20:06:28
|
On Friday 22 April 2005 03:54 pm, Mike Christie wrote: > Have you run the current 4.0.X.Y linux-iscsi? If you do > > rc.iscsi start > rc.iscsi stop > rc.iscsi start > > Does this work for you? As far as I remember, that didn't work for us. We've always ran 4.0.x.y linux-iscsi, and took us a bit of time to figure out what was going on. > Have you tried the open-iscsi/5.0.0.X tree Not yet, with the amount of problems we have had with linux-iscsi, and finally arrived at a semi-stable setup, we decided we'll wait a bit before experimenting with another implementation of iscsi driver. However, on a side note, we are trying to implement an environment where we use linux software raid 1 across multiple iscsi targets. It tooks us a bit of figuring out how to properly do it with the linux-iscsi, and we're still not happy with the results. The question would be, does anybody else use raid1 across iscsi targets, and what are your comments regarding the setup? Of course the basic functionality we are looking for is ability of the driver to detect failed devices and report it after X amount of time to the scsi layer. The second and also very important issue is being able to re-attach an iscsi target without stopping the entire iscsi interface on the initiator. In a scenario where iscsi targets are available to the initiator, then raid1 is laid over them, then lvm, then a filesystem, and multiple iscsi targets are attached to the initiator, any interruption in the existing connections is fatal. I guess if anybody has any comments/thoughts on this subject, please share them with me. thank you, dom |