we have 2 x Cisco 3750 configured with portfast 1000 full duplex to avoid spanning tree negotiation.
Chris
________________________________ From: Jeff Mohler speedtoys.racing@gmail.com To: Cristian Rojas intipu@yahoo.com Cc: Jack Lyons jack1729@gmail.com; "toasters@mathworks.com" toasters@mathworks.com Sent: Saturday, November 22, 2008 2:21:08 AM Subject: Re: cf takeover hungs Microsoft Windows Clusters
Two minutes eh? Sounds a lot like ARP issues. How is portfast/etc set on the switch ports?
Sent from my iPhone
On Nov 21, 2008, at 5:33 PM, Cristian Rojas intipu@yahoo.com wrote:
Hi Jack, we are using 2 standalone interfaces for iscsi, and a team of 2 interfaces for intra and cluster traffic.
thanks
Chris
________________________________ From: Jack Lyons jack1729@gmail.com To: Cristian Rojas intipu@yahoo.com; toasters@mathworks.com Sent: Saturday, November 22, 2008 1:21:00 AM Subject: Re: cf takeover hungs Microsoft Windows Clusters
We have had several clusters that have worked fine over take overs (just went through 6 panics after upgrading to 7.2.6!)., but since you mention igroups - I assume you are using iscsi for the luns. We are using fcp.
I assume you are using 3 seperate network interfaces for iscsi traffic and the intra cluster and public traffic?
Jack
On 11/21/08, Cristian Rojas intipu@yahoo.com wrote:
Hi guys, im seeing this behavior on different MS windows clusters attached to our FAS960, the most weird part is that i see the active node of the windows cluster connected to the target portal, but from the filer i dont see the igroup logged in for almost 2 minutes and thats when the windows clusters hung because they cant mount the quorum lun. All these while doiing a cf takeover.
I talked to probably 10 Netapp support engineers and no one can explain it.
Has anyone seen anything like this before? I have installed Host Utilities 5 and checked the windows registry to make sure the values have been tuned.
Thanks
Chris
what about the pagp and lacp and jumbo packets. I agree it sounds like a networking issue.
Cristian Rojas wrote:
we have 2 x Cisco 3750 configured with portfast 1000 full duplex to avoid spanning tree negotiation.
Chris
*From:* Jeff Mohler speedtoys.racing@gmail.com *To:* Cristian Rojas intipu@yahoo.com *Cc:* Jack Lyons jack1729@gmail.com; "toasters@mathworks.com" toasters@mathworks.com *Sent:* Saturday, November 22, 2008 2:21:08 AM *Subject:* Re: cf takeover hungs Microsoft Windows Clusters
Two minutes eh? Sounds a lot like ARP issues. How is portfast/etc set on the switch ports?
Sent from my iPhone
On Nov 21, 2008, at 5:33 PM, Cristian Rojas <intipu@yahoo.com mailto:intipu@yahoo.com> wrote:
Hi Jack, we are using 2 standalone interfaces for iscsi, and a team of 2 interfaces for intra and cluster traffic.
thanks
Chris
*From:* Jack Lyons <jack1729@gmail.com mailto:jack1729@gmail.com> *To:* Cristian Rojas <intipu@yahoo.com mailto:intipu@yahoo.com>; toasters@mathworks.com mailto:toasters@mathworks.com *Sent:* Saturday, November 22, 2008 1:21:00 AM *Subject:* Re: cf takeover hungs Microsoft Windows Clusters
We have had several clusters that have worked fine over take overs (just went through 6 panics after upgrading to 7.2.6!)., but since you mention igroups - I assume you are using iscsi for the luns. We are using fcp.
I assume you are using 3 seperate network interfaces for iscsi traffic and the intra cluster and public traffic?
Jack
On 11/21/08, Cristian Rojas <intipu@yahoo.com mailto:intipu@yahoo.com> wrote:
Hi guys, im seeing this behavior on different MS windows clusters
attached
to our FAS960, the most weird part is that i see the active node of the windows cluster connected to the target portal, but from the filer
i dont
see the igroup logged in for almost 2 minutes and thats when the
windows
clusters hung because they cant mount the quorum lun. All these
while doiing
a cf takeover.
I talked to probably 10 Netapp support engineers and no one can
explain it.
Has anyone seen anything like this before? I have installed Host
Utilities 5
and checked the windows registry to make sure the values have been
tuned.
Thanks
Chris
-- Sent from Gmail for mobile | mobile.google.com http://mobile.google.com
I assume you're using VIF on the filer(s)?
Are all the LUNs on a single filer? Multiple filers?
Are the filers using two different switches for their active interfaces within the iSCSI VIF?
Do you have multiple interfaces defined for iSCSI on the filer?
Are you using iSCSI multipath? NetApp or MS? iSCSI version?
A picture is worth 1000 words - there are so many possibilities here...
Could be bad switch trunking (clients connected to one switch, filer connected to another with no path), could be bad filer config (can you ping during failover?), could be lots of other things.
________________________________
From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Cristian Rojas Sent: Friday, November 21, 2008 9:36 PM To: Jeff Mohler Cc: Jack Lyons; toasters@mathworks.com Subject: Re: cf takeover hungs Microsoft Windows Clusters
we have 2 x Cisco 3750 configured with portfast 1000 full duplex to avoid spanning tree negotiation.
Chris
________________________________
From: Jeff Mohler speedtoys.racing@gmail.com To: Cristian Rojas intipu@yahoo.com Cc: Jack Lyons jack1729@gmail.com; "toasters@mathworks.com" toasters@mathworks.com Sent: Saturday, November 22, 2008 2:21:08 AM Subject: Re: cf takeover hungs Microsoft Windows Clusters
Two minutes eh? Sounds a lot like ARP issues. How is portfast/etc set on the switch ports?
Sent from my iPhone
On Nov 21, 2008, at 5:33 PM, Cristian Rojas intipu@yahoo.com wrote:
Hi Jack, we are using 2 standalone interfaces for iscsi, and a team of 2 interfaces for intra and cluster traffic. thanks Chris
________________________________
From: Jack Lyons jack1729@gmail.com To: Cristian Rojas intipu@yahoo.com; mailto:toasters@mathworks.com toasters@mathworks.com Sent: Saturday, November 22, 2008 1:21:00 AM Subject: Re: cf takeover hungs Microsoft Windows Clusters We have had several clusters that have worked fine over take overs (just went through 6 panics after upgrading to 7.2.6!)., but since you mention igroups - I assume you are using iscsi for the luns. We are using fcp. I assume you are using 3 seperate network interfaces for iscsi traffic and the intra cluster and public traffic? Jack On 11/21/08, Cristian Rojas < mailto:intipu@yahoo.com intipu@yahoo.com> wrote: > Hi guys, im seeing this behavior on different MS windows clusters attached > to our FAS960, the most weird part is that i see the active node of the > windows cluster connected to the target portal, but from the filer i dont > see the igroup logged in for almost 2 minutes and thats when the windows > clusters hung because they cant mount the quorum lun. All these while doiing > a cf takeover. > > I talked to probably 10 Netapp support engineers and no one can explain it. > > Has anyone seen anything like this before? I have installed Host Utilities 5 > and checked the windows registry to make sure the values have been tuned. > > Thanks > > Chris > > > > -- Sent from Gmail for mobile | mobile.google.com
-
Cristian Rojas -
Glenn Walker -
Jack Lyons