Demonstration – Configuring a ZFS Storage Appliance
Oracle ZFS Storage Appliances – Demonstration Configuring a ZFS Storage Appliances (Part 5 of 6)
>> Mick: We’re going to begin our demo now. We’ll switch to the VirtualBox Appliance here.
I’ve effectively installed the box, connected to it from the system console and switched it on. It’s booted. If you watch carefully, you’ve probably seen that it is in fact really a Solaris system and it’s waiting for me to begin configuration.
If we press any key to start – and I’ll have to fill in the basic configuration here as you can see. The reason that configuration already exists is because the system uses the dynamic host configuration protocol to go off and get network details which I want to change to be static. So I’m going to change the host in there.
The system actually has a very useful command line interface as well as a nice browser based interface. Although most of the rest of the demo will be via the browser I would hopefully spend a minute or two showing you commands that I can use from the command line as well. In fact, you can even do scripting if you’re doing lots of repetitive commands each step. Just like any computer system, you can create scripts to have them executed.
All the other details here are okay. I’ll enter a password and then esc-1 which tells the system I’ve finished entering the configuration data.
There we go. The system’s come back to me now and it’s telling me that I should talk to the appliance on the network address I’m assigned and then port number 215. I could log in here on the console and I could do things here at the command line but it’s a bit nicer to show it through the basic user interface. Certainly the early stages of the configuration are also a little bit easier doing that.
So I’ll start up a web browser and I’ll go into the interface which my system rightfully doesn’t trust. I’ll proceed anyway. And there we go. I’m connected up to the Storage Appliance. Log in as root, give the password.
The system realizes that this is the first time that machine has been used and I need to do a little bit further configuration before I can get started.
I click on start and it will take me through a half a dozen steps to get the thing up and running. First of which is to configure networking. Bear in mind that this is initial configuration and if I make a mistake or change my mind or I don’t really want to do any configuration at the moment, I can always come back a little bit later and do this.
The thing has got two built-in Ethernet interfaces. A normal physical machine we have four. One of which I configured in the early stages of configuration, which you saw earlier. There’s nothing to change here really other than I might want to make these things a little bit less anonymous.
For example here, I’ve got my data link which is my physical link and here I’ve got my interface which is the link with the IP address reside, etc.
Imagine I wanted to edit this. If you point to the object can you see there’s a trash can symbol so I could remove the information or there’s a little pen which I could click on to make changes to edit. The name of the host is hoki and I want to call it hoki datalink because I’m going to use that interface as a published interface for people access the shares on this device and reserve the other one for management purposes.
Notice while we’re on this form that I can actually use it to create link aggregations. A link aggregation is fairly sophisticated mechanism for failover and load spreading, so I can combine a number of Ethernet interfaces into one apparent physical interface. LACP aggregation then would ask me if I clicked on that to enter some information about the link aggregation.
The LACP is link aggregation control protocol and once the aggregation is up and running, the protocol will be used to inform the switch that the link aggregation exists and hopefully the switch will configure itself accordingly. It is worth noting that you need something like a layer 3 switch or certainly a network switch that supports link aggregations, which I don’t want to do. All I need to do is change the datalink. But I can also set up a VLAN.
My interface also – it just started up a little bit – I’ll call it hoki data-yf and at this layer of the protocol note I can use IP multipathing if I want to. Multipathing is not quite sophisticated as link aggregations, but it’s a very well used technology in Solaris and it provides failover and a certain degree of network load spreading. At the moment I don’t want to use that so I’ll bypass it. Just change the name. Once I’ve completed this step, just click on commit.
I didn’t apply so it’s warning me. Now I must know my DNS domain which I actually already put in at the early configuration stage.
Manny wants to know NTP, which is network time protocol. I can give it some machine name or IP addresses to go and look at to configure the network time protocol, which is very important for a system like this to make sure if it’s a storage device then the time stamps, etc. on the files that you’re recording are absolutely correct. I’m going to use local ones here.
It’s always best to have two or three at these machines.
Two will do us today. The more the merrier really, because the system will go to each of these machines and work out a time that it can deduce from both of them.
We’ll sync our clock and then we’ll commit. If there’s any problem contacting these systems, we’ll get an error. This will validate the path where they can access the NTP servers. We’re actually doing it, so I’m keeping my fingers crossed.
Now it wants to know name services because they could be storing user account details on some remote server. I may already have a number of user account setup and I can choose a variety of different name services. A name service is basically a store of administration information, perhaps IP name and addresses which specifically be DNS or usernames, passwords, IDs, group, name and password IDs, and a whole variety of other administration that will be very useful to store centrally.
Quite common ones these days is Active Directory which is a Microsoft mechanism. This client can authenticate through – if you as a user try and access a share on here then the system can go off to an active directory box and say, “Is this a valid user?” We don’t want any of these things at the moment.
Now we have to configure the actual hardware.
You’ll see a number of screens as you use this software. You’ll see things like available pools with a plus (+) and you click on the plus (+) and you can start adding things. You have to give the ZFS pool a name. The pool will be an area of space that is given to us by assigning a number of disks to it. It’s sort of a logical [9:50 inaudible], if you like. I’m going to call it some pool name. I’ll use superior which is that very large pool in North America and obviously a good description of the product. The two things go well together.
I’ve got 15 disks on this appliance. Remember it’s a virtual appliance so the capacity is actually quite small. But for demo purposes, it’s quite good. I might create a pool using all the disks depending on how I’m going to configure them shortly or I could create different pools for different system tiers.
Obviously, if you’re interest into making sure the management is simple then just create a single pool. I’ll choose a number of disks fairly randomly. Obviously, you normally think about this a bit and commit to it.
This screen is asking me for the disk storage I’ve selected, what type of data profile do I want to use? Do I want it mirrored, single parity RAID, double parity RAID, striped which doesn’t give us any redundancy but very high performance, or triple mirrored? And you can see the characteristics over here. Availability would be the redundancy so tripled mirrored would be very much available as with double parity RAID. Mirrored also provides good redundancy and also a very good performance.
There is obviously some thought required to work out exactly – for the workload you got, what would be the most suitable? I don’t want to spend too much time looking at these things now because I’m more interested in showing you the general purpose facilities of the device. So I’m going to choose mirrored.
There’s the summary of my data profile with my single pool.
>> Dave: Mick, to this pool notion, there is a question in the queue about should everything go on a single pool? Does it have to be in a single pool?
>> Mick: No. You can have as many as you like, Dave. It’s just administratively, it works out to be easier if you have a single pool because then you have a totally transparent amount of space and within that space you can add individual shares, you can create different projects. So really, you don’t need more than one pool unless you want to have a completely different type of data profile there from one type of server to another. But usually speaking, you just have a single pool.
>> Dave: Perfect. Thanks, Mick.
>> Mick: Okay. The GUI is now allowing us to register online, which we don’t need to do at the moment.
Copyright SkillBuilders.com 2017