Re: GFS and Qmail and BIG mail servers
Also have a look at silicon-gear.com, that also makes a Dlock array. -jeremy On Fri, Jan 28, 2000 at 12:41:10AM -0800, Tracy R Reed wrote: I have just become aware of the GFS project and I am BLOWN AWAY. I don't know how this project reached production quality status and escaped my radar until now. I got an email from my VAR for StorageTek disk arrays today GFS is most cool. You should be aware, however, that it is not yet production quality. The journaling code is mostly done, but it is not stable yet. Client errors can (and do) result in corruption of the entire filesystem. (As of a few weeks ago) It should be nearing Alpha quality at this point, but it is not ready for production use. and tell me what you think. Does this sound like a good way to cluster mail servers to you too? It is the holy grail of mail clustering. :) It's just not there quite yet. I have priced it out, though, and you can build quite an effective (and large) mail cluster for under 100k with it. If you're still interested in looking at it, BoxHill makes some excellent storage arrays that support hardware SCSI Dlocking. Good juju. Adam -- -- Adam Jacob - Cyber TrailsPhone - (602)906-1752 Sr. Systems AdministratorPager - (602)447-9531 [EMAIL PROTECTED] Fax - (602)906-1799 * Evil Lord of the Sysadmin Sith Darth Rmdashrf * -- http://www.xxedgexx.com | [EMAIL PROTECTED] -
GFS and Qmail and BIG mail servers
I have just become aware of the GFS project and I am BLOWN AWAY. I don't know how this project reached production quality status and escaped my radar until now. I got an email from my VAR for StorageTek disk arrays today pitching it as a solution. This appears at first glance to be the answer to a lot of our mail server scalability problems. You can now put a bunch of SMTP and POP servers (DNS round robin or serveriron/bigip them) clustered on a fibrechannel network talking to shared disk arrays. No NFS, no TCP, just SCSI over fibrechannel with every host in the cluster talking to every disk. I just finished reading their latest IEEE paper and the locking mechanism makes sense and seems quite sane and suitable to handling just this sort of situation. It's a journalled and fault tolerant filesystem. If a machine in the cluster goes down, no problem. With well supported and relatively inexpensive Qlogic fibrechannel cards and fibrechannel disks costing only slightly more than SCSI disks (I think, I've never really priced them individually) this seems quite viable. I encourage all of you to check out: http://www.globalfilesystem.org and tell me what you think. Does this sound like a good way to cluster mail servers to you too? -- Tracy Reed http://www.ultraviolet.org Every one we don't catch would be a "yet another major ms security hole", and the theory tells us we can't catch all of them. So, we're just not going to start down that path. [EMAIL PROTECTED] 08/06/98 Bugtraq
