Mailing List Archive: Netapp: toasters

SD 6.01 / W2K8 x64 / iSCSI and NLB Cluster

Index | Next | Previous | View Threaded

phigmov at gmail

Nov 28, 2008, 11:34 AM

Post #1 of 3 (2205 views)
Permalink

SD 6.01 / W2K8 x64 / iSCSI and NLB Cluster

Hi all,

A few queries about the above config -

We are setting up a new E2k7 system with 2 front end CAS servers with
NLB and 2 back end Mailbox servers in NLB config and doing CCR between
them (no shared storage).

I was able to install SnapDrive 6.01 and create four LUN's (non shared
dedicated) on each Mailbox server on two seperate SAN's for the mail
stores & logs. No problem.

However as soon I rebooted and went to check the LUN's (which stayed
connected) and create a new one I got a couple of errors (different
on each system)
- RPC server unavailable
- Security package related error

Oddly I can expand existing LUN's OK via SD - so some functionality
exists - just can't create a new LUN within SD.

I have had a suggestion to make sure the Windows Quorum Witness Disk
should also be on the SAN - following Microsoft best practive I've
created a small Quorum Witness Disk and share on a CAS server. Thats
fine - still doesn't fix the RPC error (I can ping and access the
admin share on the SAN from each mailbox server).

The most annoying thing is that none of these issues has cropped up
with W2K3 and SD 4.2.1 - I'm following the same fundamental setup
steps.

Given I'm not using shared storage clustering just CCR and NLB I want
to discover the root of the SD errors
- is it SD trying to be clever and insisting that for a Windows
cluster you really need shared disks registered with the cluster
administrator regardless of wether they're shared or not ?
- just some weirdness with W2K8 and SD 6.01 ?
- are there any gotchas with W2K8 and SD 6.01 that I need to be aware
of ? I just noticed theres a DCOM exception for the W2K8 firewall on
NOW - how come SD doesn't set this stuff up at install time ?

I want to eliminate the clustering ASAP issue because until I nail the
problem the Exchange guys can't really proceed with the rest of the
install. Particularly if I need to keep messing around with breaking
and re-establishing the cluster to sort out an RPC error.

Thanks in advance,
Raj.

phigmov at gmail

Dec 2, 2008, 1:46 PM

Post #2 of 3 (2081 views)
Permalink

Re: SD 6.01 / W2K8 x64 / iSCSI and NLB Cluster [In reply to]

Cheers Tom

We've done some more troubleshooting -

* if I do a clean w2k8 server install with two nics (one data and the
other iscsi) - SD 6.01 installs and works fine - the initiator even
prompts to open the FW ports on the server

This made me think it was either something fishy with my previous
install or something funny with the cluster (which is purely failover
- no shared storage).

* if we evict one of the cluster members (there are two mailbox
servers) the SD 6.01 RPC problem goes away immediately

So now I'm running up some test VM's to replicate the setup and run up
SD first and then put the cluster on afterwards.

What I'm wondering now is what part of the SD / SAN interaction is
making the RPC call - presumably over the clustered address ? Is
there any way to tie it to a particular IP or NIC ?

We'll run another test with a couple of vm's, creating the luns and
then creating the cluster - see at what point the RPC error occurs.

Cheers,

Raj.

On Tue, Dec 2, 2008 at 10:04 PM, De Wit Tom (Consultant)
<tom.de.wit [at] consultant> wrote:
> I know of two things that can cause such errors:
>
> - Fill in the "Preferred filer IP addresses". I don't have SD 6.x, on 4.2.1 it was located in MMC under Storage/Snapdrive/Disks, rightclick Disks, select properties. Then fill in the filer DNS/Netbios name and the corresponding IP address and restart Snapdrive services
>
> - Make sure the Snapdrive services is started with a domain user that has both local admin rights on the local server and admin rights on the Netapp filer.
>
> Grtz,
> Tom
>

phigmov at gmail

Dec 3, 2008, 4:26 PM

Post #3 of 3 (2065 views)
Permalink

Re: SD 6.01 / W2K8 x64 / iSCSI and NLB Cluster [In reply to]

Cheers Tom,

We get slightly different errors on the nodes -

* rpc error - mailbx1
* security package error - mailbx2

We did have a suggestion to completely disable the firewall service
(not just switch it off in the control panel) - this worked until we
rebooted both nodes and the problem came back.

Why I asked about the NIC's was because I wondered if the RPC traffic
was going out over the data network rather than the iSCSI network. The
data network potentially has two addresses per host - its own and the
failover cluster IP.

I don't think we're doing anything particularly fancy - the cluster
setup is obviously the complicating factor. There are a few NOW
articles concerning cluster setup but they tend to relate to
traditional shared disk rather than independent disk setups.

Cheers,
Raj.

On Wed, Dec 3, 2008 at 11:00 PM, De Wit Tom (Consultant)
<tom.de.wit [at] consultant> wrote:
> Hey Raj,
>
> Setting the Preferred filer address is what will tie those RPC calls to a specific destination IP address. Normally you do this to make that traffic go over a specific source interface. The source IP address woudln't matter that much if the traffic can get back to the same host using that IP address.
>
> Are you experiencing these SD errors on both of your cluster nodes ?
>
> Tom
>
> -----Original Message-----
> From: Raj Patel [mailto:phigmov [at] gmail]
> Sent: dinsdag 2 december 2008 22:47
> To: De Wit Tom (Consultant)
> Cc: toasters [at] mathworks
> Subject: Re: SD 6.01 / W2K8 x64 / iSCSI and NLB Cluster
>
> Cheers Tom
>
> We've done some more troubleshooting -
>
> * if I do a clean w2k8 server install with two nics (one data and the
> other iscsi) - SD 6.01 installs and works fine - the initiator even
> prompts to open the FW ports on the server
>
> This made me think it was either something fishy with my previous
> install or something funny with the cluster (which is purely failover
> - no shared storage).
>
> * if we evict one of the cluster members (there are two mailbox
> servers) the SD 6.01 RPC problem goes away immediately
>
> So now I'm running up some test VM's to replicate the setup and run up
> SD first and then put the cluster on afterwards.
>
> What I'm wondering now is what part of the SD / SAN interaction is
> making the RPC call - presumably over the clustered address ? Is
> there any way to tie it to a particular IP or NIC ?
>
> We'll run another test with a couple of vm's, creating the luns and
> then creating the cluster - see at what point the RPC error occurs.
>
>
> Cheers,
>
> Raj.
>
>
> On Tue, Dec 2, 2008 at 10:04 PM, De Wit Tom (Consultant)
> <tom.de.wit [at] consultant> wrote:
>> I know of two things that can cause such errors:
>>
>> - Fill in the "Preferred filer IP addresses". I don't have SD 6.x, on 4.2.1 it was located in MMC under Storage/Snapdrive/Disks, rightclick Disks, select properties. Then fill in the filer DNS/Netbios name and the corresponding IP address and restart Snapdrive services
>>
>> - Make sure the Snapdrive services is started with a domain user that has both local admin rights on the local server and admin rights on the Netapp filer.
>>
>> Grtz,
>> Tom
>>
>

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads