4530 Network config

[AngusMacdonald]

A few years ago, a 3rd party supplier installed and configured a single 4530 for us as part of a larger project. Unbeknown to the techies at the coal face, the contract stipulated that we were responsible for supporting all the kit and the supplier did not (and now cannot) provide any handover documentation. We are now experiencing some issues and I have the job of dealing with them. This is the first time I've looked at storeVirtual and the system it supports is critical, so no pressure. Two clustered Windows servers connect to the 4530 but at the moment only one of them can connect to the LUNs. After a reboot for Windows updates, the other shows LUNS stuck in "Connecting" state.

The unit has  three NICs configured.
FlexLOM:Port1 is statically configured with an IP address and mask, but no gateway on a management VLAN
NICSlot2:Port1 is statically configured also with an IP and mask, but no gateway, on a storage VLAN
NICSlot2:Port2 is statically configured with IP, mask and gateway on the storage VLAN
There are no bonds configured
The Lefthand OS interface is configured on NICSlot2:Port1 (ie no default gateway)
The CMC can only detect the 4530 from a machine on the storage VLAN. It's undetectable from any routed VLAN

So, to my questions:
Should both storage VLAN NICs have default gateways?
Should the Lefthand OS Interface use a NIC with a gateway?
Can the lefthand OS interface use the  FlexLOM:Port1 NIC on the management VLAN? The info on this page  suggests the OS interface should use the storage interface.
And if so, should I presume that management interface should have a default gateway?
I take it that bonding the two NICSlot2 ports is strongly recommended?
Making any TCPIP config changes (even on the management interface or assigning a new lefthand OS NIC) displays a dire warning about downtime. Which actions will actually prompt unavailablility?

Sorry for all the (undoubtedly newbie) questions but the criticality of the system (I work in healthcare) and my lack of experience with StoreVirtual makes me extremely nervous!

Angus

[AngusMacdonald]

Everything has now gone down. None of the servers can mount any LUNs, although the LUNs themselves are visible to the iSCSI initiator. Since all is now broken I've gone ahead and configured a bond on the two 10G interfaces. The servers can see their LUNs with the bond address as the discovery address but they still cannot connect to them. Should the initiator be configured to discover on the bond address or the cluster VIP address? The latter doesn't work but perhaps the VIP address is somehow wrong.

Update: The dicovered LUNs have "Associated Network Portals" (MS iSCSI Initiator) addresses matching the cluster VIP address rather than the discovery address. Is that normal?

[AngusMacdonald]

A further update. I spent the night creating 22 new LUNs on our Equallogic infrastructure and restoring to them from Arcserve. The 4530 is now languishing in the bin.

Lesson learned about allowing DXC to dictate contract terms.

[kannoly]

Hi Agnus, 

There is no need for Gateway. 

Only recommendation is that for management you need to use eth0 . LHOS and ISCSI network can use other interface. 

Most of the customer use Management on eth0 and create a bond for LHOS and ISCSI .

If you are creating a two bonds, make sure  that you assign bond 0 for manageent and bond 1 for ISCSI and LHOS. 

For your issue, you disconnect the ISCSI session from the host which is not accessible and reconnect the session to the VIP

It should get detected 

Regards

Vipin 

[kannoly]

The initiator should always connect to the Virtual IP address.  This will help failover.

When a storevirutal nodes goes down, VIP will be moved to a different node in the cluster  automatically.  Thus keeping  all the iscsi sessions alive.

[kannoly]

The initiator should always connect to the Virtual IP address.  This will help failover.

When a storevirutal nodes goes down, VIP will be moved to a different node in the cluster  automatically.  Thus keeping  all the iscsi sessions alive.

[kannoly]

Hi Agnus, 

Please let me know if you need any help from the  StoreVirtual end.  I can help you to fix store Virtual issue. 

Regards

Vipin 

[kannoly]

You can also refer the below advisory 

https://support.hpe.com/hpsc/doc/public/display?docId=emr_na-a00019438en_us 

[AngusMacdonald]

Thanks Vipin.

The SAN was configured as you describe it, to the best of my knowledge. iSCSI clients could not connect to the VIP but could connect to the Bond address and could "see" their LUNS via that route. However, they could not connect to the LUNs, giving an error every time. All three affected servers and the SAN itself were subject to power-off reboots but nothing helped. At that point, with my hospitals patient admin system down at 1am, I decided to abandon the 4530 to its fate and migrate to our Equallogic infrastructure. We won't be returning to StoreVirtual. In truth, we shouldn't have used it in the first place but our contract with DXC was weak.

Angus