napp-it all in one - newbie questions

[vjeko]

OK I'm finally back and still at newbie level

I wanted to change the omnios ip address , so I saw that it was possible in napp-it. After changing the ip address, I could not get napp-it to start-up in the web browser. In esxi in the omnios console, the new ip address is listed but it has a /8 network mask whereas my network is using /24 - not sure if this is the problem / how do I get napp-it to work again ?

 

[gea]

I would first check the current ip settings at console with the command
ifconfig -a

Try to connect to the displayed ip
If this is not possible, you can
- go back to a former BE (select at bootup) or
- delete the link manually and enable dhcp or set a new ip manually

like it is done on an initial setup of napp-it, see
napp-it // webbased ZFS NAS/SAN appliance for OmniOS, OpenIndiana, Solaris and Linux : OmniOS

In my own setup, i mostly use the e1000 for management and the vmxnet3s for ESXi/data.
In case of problems you have the second ip for management

 

[vjeko]

OK, I've been more off than on - finally back.
I must have botched something up - hopefully I can fix this without
taking a Unix course (the knowledge from which would be more than desirable )

First, please comment whether ESXI and OmniOS should have static
ip addresses (I had them static but then opted for setting the MAC/IP/host names in the router
and setting both ESXI and OmniOS as DHCP.

What do you mean by "former BE" ? (todate, I've just fiddled around
not making any backups or anything yet)

-If I do a reboot of Omnios, there are two options:
-default
-backup_151014_Sep_2015
Can't remember back to install, I guess the backup came automatically
with the install. I use the "default" boot option for OmniOS.

For some reason I understand the physical/virtual components,
but still (specially after beinng on/off this) have some problem
understanding the vm network and management network
and assignment of the vnics - if you don't mind a few details
about where you added the e1000 for management and the vmxnet3s for ESXi/data ?
For the physical nics, at the moment I've got just
vmnic0 (inbuilt nic) connected and have
the ibm quad nic for future needs - so both the communication
to ESXI ip via vsphere client as well as ping to OmniOS
are going via that one nic.

After fiddling around with the "delete addr" command and going through the steps
for creating a new DHCP address on e1000g0,
-"ifconfig" gives the correct ip address (and netmask - "/24" i.e. "ffffff00") for OmniOS
as set in the router MAC/IP list but 127.0.0.1 netmask is "ff000000" (mistake from before).
-I can ping the new OmniOS ip address from the pc which is running the vSphere client
- can't reach napp_it via OmniOS_ip:81

How can I check if napp_it is still installed ?

 

[gea]

ESXi networking

ESXi offers you a virtual switch that works similar to a real vlan capable switch.
One port is the uplink port and connected to the physical nic. You can provide there untagged or tagged ethernet with vlans.

Most ports of the virtual switch are switchports where you can connect virtual nics from your VMs (VM network). You can connect these vnics untagged or tagged to a vlan number.

A VM can have one or more virtual network adapters that are connected to the virtual switch. You have the choice between e1000 (slow but supported on any OS even without VMware tools installed) or vmxnet3 that is faster but require that you install the VMware tools.

One port of the virtual switch is the management port, similar to a real switch. In this case ESXi itself is listening on this port.

You can separate traffic between VMs, the management port or the physical network either with vlans or multiple switches that can be connected to different physical nics.

The backup boot environment
is a freeze of the downloaded state. If you select this BE you have the state after first bootup.

napp-it
You can check if napp-it is installed when you try to start (root at console)
/etc/init.d/napp-it start

 

[vjeko]

Would you then suggest one vswitch&nic for ESXI/Vsphere client
and a separate vswitch&nic for OmniOS (I'm running untagged vlans/no smart switch) ?

Benefit of having ESXI and OmniOS dhcp vs static (dhcp via router static MAC/IP list)?

OK, I tried the following on the default BE:
ipadm delete-if e1000g0
dladm show-link
ipadm create-if e1000g0
ipadm create-addr -T dhcp e1000g0/dhcp
reboot
(didn't issue the nameserver,dns or napp-it commands as they were done
before)
ifconfig
- shows IP address as per router MAC/IP list for OmniOS
but netmask for ip address is ffffff00 (correct), but for 127.0.0.1 is ff000000 (wrong)
-ping from pc running Vsphere client works towards this ip address
-OmniOSipaddress:81 doesn't work in browser


init.d/napp-it start ,results in:
server available at 127.0.0.1:3000 (but just to remind wrong netmask for 127.0.0.1)

 

[gea]

127.0.0.1 is a class-A adress so ff000000 (255.0.0.0) is ok
This is also not relevant as this is a loopback reference.

btw
napp-it is listening to all interfaces unless you activate the OmniOS firewall

 

[vjeko]

I'm no network expert, but what I was alluding to was the fact that
the netmask of the ip address (10.X.X.X) I gave to OmniOS was /24,
whereas for the 127.0.0.1 (my thinking was maybe that's why I can't
access napp-it). Still don't know how to get to napp-it -any comments
on results of what I tried/what else should I try ?

 

[gea]

Depends on your concrete settings but 10.0.0.0 (/8) is also a class-A private network.
Maybe your DHCP settings are the problem (ip adresses not in the same /24 range).

Private network - Wikipedia, the free encyclopedia

Try to use adresses from 192.168.1.x in a /24 (255.255.255.0) subnet mask.

ex dhcp from 192.168.1.100 to 192.168.1.254, netmask 255.255.255.0 (/24)
(dhcp adresses must exclude router address but must be in the same ip range)

 

[vjeko]

OK, looks like I've been sloppy regarding abiding by the netmask rules for 10.X.X.X, but
it has worked with a lot of devices hooked up (smartphones, pc's, laptops , ESXI server and OmniOS/napp-it). Anyway, I'll change it tomorrow morning to be according to spec and give it a try
by removing /adding the link again.
Is booting the backup BE the only other way to get to napp-it in case this doesn't work ?

I guess the BE's are not of vital importance as all data/settings are on the storage disks, but what's
the usual procedure regarding BE's - should I have made some backups immediately
after install or is reinstalling the VM the answer ? I ask because I would like to go through disaster recovery for each of the scenarios (simple as above,OmniOS disk dies, data disk(s) die ) before I can get comfortable with storing a lot of data on the server.

I would really like a comment regarding benefit of DHCP (via router) vs static ip address for ESXI and OmniOS(I'm too green in this field - I see the benefit of static but can't see how DHCP would
cause problems )

 

[gea]

Boot environments is a good way to go back to a state prior an update or a critical modification. For an AiO this is not too important as even a complete reinstall is done in a short time.

regarding manual ip
As you should enter the ip adress of your NFS storage in ESXi, your storage must keep this ip. With DHCP it can happen that you get a new ip if you power off a system for some time. You should not use a hostname as this require a DNS server.

Yiu can use dhcp then but only with a fix ip assignement there based on MAC adresses.

 

[vjeko]

Can the BE be recreated ?
Now to simulate the disaster situation - do I need to do anything special (or are there any
special instructions) during or after a OmniOS re-install (as I said, I want to be sure I
can do these things before I have faith in the system )

Yes, all my ip's are static MAC/IP (dhcp on router gives out the same ones always).
Looks like the router is then a DNS server as ESXI picked up the hostname + domain name
stored in the router.

 

[gea]

A ZFS BE is a snap + cloning mechanism so you can boot from.
You can create and destroy them when needed. They are a perfect undo mechanism for system changes.

Worst case if a bootdisk with ESXi and napp-it fails completely:
Reinstall ESXi (10 Min), import the ZFS napp-it storage VM (3 min), set up vt-d for the HBA (2 min inkl. ESXi reboot). Now you can import your datapool with the NFS share already enabled. Add the NFS to ESXi as a datastore.

Open the Vsphere file browser, open a VM folder. Add the VM with a mouse-right click on the .vmx file, add to inventory. Boot your VMs. You are asked if the VM is cloned or copied, answser copied. (1 min per VM)

 

[vjeko]

If possible, I would like to learn a bit more from this problem than just being able to use
the other BE as I was merely changing ip addresses / static/dhcp and deleting links,
so I would like to know how to get napp-it working after such changes.

I viewed Omnios files (Windows-> Midnight Commander via putty)
and saw a file /etc/rc3.d/S99napp-it. There the listed ip is an old one, so maybe I've
missed doing something for napp-it after the ip change or do I need to re-install napp-it when
an ip is changed / what else should I try ? If I do re-install napp-it again, do I need
to delete anything beforehand ?

 

[gea]

If you install napp-it via wget or update/downgrade with the GUI, all files like the startup file are updated.

If you mess up the network, you can delete/recreate the interface similar to the initial setup. But there are other that are more complicated to fix. This is why you should care about BE's as bootable system snaps that you can create regularly or prior critical system changes.

Without a BE a reimport of the napp-it template with a pool import is mostly faster than any effort of repairing the OS as this is done withing 2 minutes unless you have not installed extra services with a complicated setup. In this case, you should think of backup ex saving the current state as an ESXi template than can be restored also within 2 minutes.

 

[vjeko]

Attached is a photo of all BE's :
-first 2 BE's are from initial install
-3 new BE's from issuing "wget -O - www.napp-it.org/nappit | perl" using the
"default" BE
- "160809_with_pw" - snapshot of BE created from "backup_151014_Sep_2015_"

What I did:
(a)Using "default" BE and issuing "wget -O - www.napp-it.org/nappit | perl",
the 3 new BE's were created . I tried rebooting to one of the new BE's but again I couldn't
access napp-it via web.
(b)Using "backup_151014_Sep_2015_", Omnios was without password / I could
log into napp-it via web.

Questions:
(a)I want to keep BE's :
-"backup_151014_Sep_2015_"
-"160809_with_pw"
which BE's can I delete ( I see that "default-1" can't be marked for deletion) ?

(b)To get the pools back, I guess I do "Pools->Import" without any options set ?

 

[gea]

a
you can delete all BE beside
- the current (N)
- the default that is used on a reboot if you do not select one manually (R)

but care about:
A BE is not always the data state on creation time. This is only true for a backup BE that you never boot.
If you boot into a BE, all data modifications that you are doing then are part of that BE then. This is the reason why a quite old BE that is your default BE has an old creation time but is the most up to date BE.

b
yes