Hello everyone, I will be upgrading my filer to redundant fcals this week and I wanted to check to see if anyone is already doing this. I don't think there will be any tricks to it, but it never hurts to ask.
Best Regards,
Josh J Gifford MCP Senior NT Administrator Siemens Power Transmission & Distribution 7000 Siemens Rd Wendell, NC 27591 Phone: (919) 365-2806; Fax: (919) 365-1080 Email: josh.gifford@ptd.siemens.com
We're not actively using it, but we tested it. It was pretty darn slick! Its kinda bizarre though because "vol status -r" output shows alternating disks on alternating adapters - until you failover to one or the other, then they all show disk id's on the active adapter. A load balancing strategy I assume, but it made us double take when looking at the output! =)
-- Jeff
On Mon, Apr 23, 2001 at 12:50:05PM -0400, Gifford, Josh wrote:
Hello everyone, I will be upgrading my filer to redundant fcals this week and I wanted to check to see if anyone is already doing this. I don't think there will be any tricks to it, but it never hurts to ask.
Best Regards,
Josh J Gifford MCP Senior NT Administrator Siemens Power Transmission & Distribution 7000 Siemens Rd Wendell, NC 27591 Phone: (919) 365-2806; Fax: (919) 365-1080 Email: josh.gifford@ptd.siemens.com
jkrueger@qualcomm.com (Jeffrey Krueger) writes
On Mon, Apr 23, 2001 at 12:50:05PM -0400, Gifford, Josh wrote:
Hello everyone, I will be upgrading my filer to redundant fcals this week and I wanted to check to see if anyone is already doing this. I don't think there will be any tricks to it, but it never hurts to ask.
[...]
We're not actively using it, but we tested it. It was pretty darn slick! Its kinda bizarre though because "vol status -r" output shows alternating disks on alternating adapters - until you failover to one or the other, then they all show disk id's on the active adapter. A load balancing strategy I assume, but it made us double take when looking at the output! =)
At one time this would have got people talking about "the performance penalty of writing to discs on more than one FCAL controller", wouldn't it? But presuambly not now that Sam Rafter set us straight on the subject in his toasters posting of 12 Jan 2001... or could the scenario of burt 19290 ever apply in the context of "redundant FCALs"?
Chris Thompson University of Cambridge Computing Service, Email: cet1@ucs.cam.ac.uk New Museums Site, Cambridge CB2 3QG, Phone: +44 1223 334715 United Kingdom.
* It was Tue, Apr 24, 2001 at 12:32:24PM +0100 when Chris Thompson wrote:
jkrueger@qualcomm.com (Jeffrey Krueger) writes
We're not actively using it, but we tested it. It was pretty darn slick! Its kinda bizarre though because "vol status -r" output shows alternating disks on alternating adapters - until you failover to one or the other, then they all show disk id's on the active adapter. A load balancing strategy I assume, but it made us double take when looking at the output! =)
At one time this would have got people talking about "the performance penalty of writing to discs on more than one FCAL controller", wouldn't it? But presuambly not now that Sam Rafter set us straight on the subject in his toasters posting of 12 Jan 2001... or could the scenario of burt 19290 ever apply in the context of "redundant FCALs"?
When the second FCAL loop is in hot standby mode, or whatever term we've given it, it's generating no appreciable amount of traffic, so 19290 isn't a threat. Also, if this is an f800, it's not an issue at all. We revisited that bus design.
I retract my earlier statement. I've had an avalance of mail explaining to me how the active-active FCAL setup works, and it's not what I'd envisioned earlier. It's much cooler. When I'm sure I understand it, and my toasters mailing privaleges are returned, I'll try to answer this question again.
* It was Tue, Apr 24, 2001 at 08:49:00AM -0700 when Sam Rafter wrote:
- It was Tue, Apr 24, 2001 at 12:32:24PM +0100 when Chris Thompson wrote:
jkrueger@qualcomm.com (Jeffrey Krueger) writes
We're not actively using it, but we tested it. It was pretty darn slick! Its kinda bizarre though because "vol status -r" output shows alternating disks on alternating adapters - until you failover to one or the other, then they all show disk id's on the active adapter. A load balancing strategy I assume, but it made us double take when looking at the output! =)
At one time this would have got people talking about "the performance penalty of writing to discs on more than one FCAL controller", wouldn't it? But presuambly not now that Sam Rafter set us straight on the subject in his toasters posting of 12 Jan 2001... or could the scenario of burt 19290 ever apply in the context of "redundant FCALs"?
When the second FCAL loop is in hot standby mode, or whatever term we've given it, it's generating no appreciable amount of traffic, so 19290 isn't a threat. Also, if this is an f800, it's not an issue at all. We revisited that bus design.
-- Sam Rafter Escalations Jerk Network Appliance rafter@netapp.com
On Mon, 23 Apr 2001, Gifford, Josh wrote:
I will be upgrading my filer to redundant fcals this week and I wanted to check to see if anyone is already doing this. I don't think there will be any tricks to it, but it never hurts to ask.
What benefit does this bring, beyond protecting against an FC-AL failure on a single Netapp? Any additional benefits while in a clustering scenario?
On Mon, Apr 23, 2001 at 09:50:58PM -0400, Brian Tao wrote:
On Mon, 23 Apr 2001, Gifford, Josh wrote:
I will be upgrading my filer to redundant fcals this week and I wanted to check to see if anyone is already doing this. I don't think there will be any tricks to it, but it never hurts to ask.
What benefit does this bring, beyond protecting against an FC-AL
failure on a single Netapp? Any additional benefits while in a clustering scenario?
As long as we are talking FC-AL, DMP and clustering (on NetApps) are mutually exclusive. The disk shelves still have only two ports therefore if you cluster (one port to each host), there are no ports left to have redundant paths to either host.
If I were consulting, I'd say that if you don't want to full cost of clustering, but still want some degree of "high availability", that DMP is useful. Especially on big FC-AL loops with lots of devices, one node (disk or host) can really hold the entire loop hostage and wedge the filer. In this scenario, DMP just switches all drive communications over to the functioning path.
I'm sure there would be disagreement on this, but it may pose a slight performance increase because it appears to load-balance across the redundant HBA's if neither is failed out. Whether this is true or not, that performance probably never gets realized at the client's side due to the fact that filers aren't usually disk bound anyhow.
Oh, and I'd like to put a motion on the floor that Sam's posting authority to toasters be fully re-instated. =)
-- Jeff
-- ---------------------------------------------------------------------------- Jeff Krueger, NetApp CA E-Mail: jeff@qualcomm.com Senior Engineer Phone: 858-651-6709 NetApp Filers / UNIX Infrastructure Fax: 858-651-6627 QUALCOMM, Inc. IT Engineering Web: www.qualcomm.com