Storage network connectivity question

your goal is to have at least one SPA and one SPB port on each Cisco switch, then each host would be zoned like this:

Fabric A (Cisco switch 1)

host1-zone1-hba1-SPA0

host1-zone2-hba1-SPB2

Fabric B (Cisco swtich 2)

host1-zone1-hba2-SPB0

host1-zone2-hba2-SPA2

once the new logical paths are established you will need to go into Connectivity Status and make sure to register them (same settings as existing initiators). Once paths have been registered you will need to go to Storage group properties, hosts tab. Send CTRL+SHIFT+F12 combination and enter password "messner". At this point you will see "Advanced" bottom appear on the bottom of the Hosts tab, select that option and then make sure you put a check mark next to the new paths you just registered.  Now go back to ESXi and rescan each HBA, then go to Storage > Devices, select Manage Paths and you should see all the paths to that device.

Hi dynamox, thanks for your quick reply.

As you can see i have attached a screenshot of both our connectivity and port management current status. The reason why would like to add more physical connections from the storage system to our switches is because we would like to increase the bandwidth since we are facing some disk latency issues with our VM's. As per above, we already have 2 connections from each SP to each switch.

Adding an extra physical connection from each SP to each switch will increase the overall bandwidth or design wise we should proceed differently (assuming of course that the ESXi servers and the switches also have enough bandwidth)?

Regarding the registration of the new paths, i can identify the settings from the other initiators but i am a bit confiused regarding the WWN/IQN field. How do i find the correct information for that field? On all of my other initiators, as you can see from the screenshot, the value starts from 20:00:00:C0:DD...... When i go to Port Management and identify the newly connected port, the IQN/WWN there goes something like 50:06:01..... the same goes for my already registered initiators. So, how do i identify/calculate the correct number to use?

Our Software Version is 04.30.000.5.509 for CX4-240 if it matters.

Thanks again,

Charis

Charis,

Do you have access to Navisphere Analyzer ? The reason i ask is because you already have 16G worth of bandwidth to each LUN (let's say a LUN is owned by SPA,  your server is connected to SPA0 and SPA1). Are you really driving 16G worth of data that you feel that is your bottleneck ?  If you have access to Navisphere Analyzer i would look at response times of your LUNs, look at disk utilization first.

We didn't purchase Analyzer so unfortunately no i do not have access.

Let me just explain why we are doing this: Recently we started getting the following error on our ESXi servers: "Device naa.xxxx performance has deteriorated. I/O latency increased from average value of xxxx microseconds to xxxx microseconds". At the same time many VM's report the "Virtual machine total disk latency" alarm. This happens 5-6 times a day, VMs performance dramatically decreases for 5-10 minutes and then everything gets back to normal.

All hard drives in the storage seem to be working fine, no error reported in the storage health or ESXi health and the switches seem to be working fine - so we assume there is no obvious hardware failure. We have also upgraded our HBA firmware on ESXi's after VMware support recommendation without any success.

The only indications we found was both on the switches and the storage LOG files, at the time the problem is occurring, there is a connectivity loss for some seconds and then the connectivity gets restored. The reason we did not assume that there is a hardware failure is because we get these errors on both switches, at the same time and at the same ports (the ports that ESXi servers are connected on).

So we figured that perhaps there is a lack of bandwidth: either from the storage to the switch or either the ESXi does not have enough networking capabilities to satisfy all the VMs.

We currently have 135 VM running on the above configuration and each ESXi has 2 x 10gb FoE cards

Is there a way to measure the overall bandwidth consumed for the storage without the Analyzer or do you think we might be looking at the wrong direction?

Charis

Ok then i will open a support ticket today. I do want to ask you though, usually what are the reason where forced flushing occurs? Not enough bandwitdh / bad configuration / slow hard drives..?

Primus emc186107

Dynamox thank you very much - you've been very helpful!

Charis

no problem, let us know what you get from support.