12-10-2013 06:17 AM
Hi,
Im about to add a new node to a cluster (Solaris OS and version 5.0 for the cluster) and I have set all the cabling to have the HB network working. The heartbeat links are connected correctly using a switch over the same VLAN.
I was checking if the HB links are up but not plumed, but they are in this state unknown.
root# dladm show-dev
igb0 link: up speed: 1000 Mbps duplex: full
igb1 link: unknown speed: 0 Mbps duplex: unknown
igb2 link: unknown speed: 0 Mbps duplex: unknown
igb3 link: unknown speed: 0 Mbps duplex: unknown
usbecm0 link: up speed: 10 Mbps duplex: full
Its the interfaces igb1 and igb2.
Im new at this and comparing to the other nodes and internet examples, their HB interfaces are up but NOT plumbed...
So, these interfaces - igb1 and igb2.- should be up just by plugging the cable or should I issue a command to do so?
Is there any checnking I can do to see where the problem is, if there is any?
Tks in advance,
Joao
Solved! Go to Solution.
12-10-2013 09:48 AM
Sorry, couldn't see previous comments for some reason ... pinging on self machine is not gonna make much sense ...
Because heartbeat is already up & running in the cluster & if cluster is a production cluster, I wouldn't recommend to try plumbing an IP on heartbeat.
Your ultimate task is to add the node to cluster, so anyways you would need to install VCS packages. You can install VCS packages (download from symantec portal) with that, you will get VRTSllt package.
with VRTSllt installed & all the cluster config files (llttab, llthosts,gabtab) populated, you can then run the dlpiping or lltping tests as explained in first step
G
12-10-2013 09:56 AM
If you can't assign IPs to other node as it is live, then you cant test using IPs. There shouldn't be an issue assigning IPs to a live node from a technical point of view, but running any command on a live server can be risky if you make a mistake (hence why most live systems will have change control)
Mike
12-10-2013 06:23 AM
Hi Joao,
LLT status up is good & LLT interfaces will not be plumbed...
you can use dlpiping command or lltping or command to view the llt status ...
see this article for using dlpiping
http://www.symantec.com/docs/TECH19998
see below doc for using lltping
http://sfdoccentral.symantec.com/sf/5.0/solaris/manpages/vcs/lltping_1m.html
G
12-10-2013 06:34 AM
Tks for the fast reply.
I havent installed it yet so I dont have the veritas packages installed yet to run either of the commands you sent...
I think I will have to install it and then test right?
Is there any command like dladm to test it? I mean a unix command...
Tks,
Joao
12-10-2013 07:29 AM
To test heartbeats, I ususally just temporarily plumb them in with test addresses and use ping to test - so something like:
on node1 give IPs
1.1.1.1
1.1.2.1
and on node 2
1.1.1.2
1.1.2.2
all with netmask 255.255.0.0.
And if ping doesn't work then you know it is not cabled correctly or there is an issue with VLAN where as if you use the unfamilar dlpiping tool you won't be sure if you are using tool correctly.
Mike
12-10-2013 08:39 AM
Tks Mike.
Sorry for the newbie question...so I will plumb the interfaces in the new node with dummy IPs and try to ping them from the other node... is this what you suggested?
Btw, the interface be up and not plumbed for the Veritas installation, right? And this up of the link should be automaticallly when you plug the cable or a command is necessary?
Tks,
Joao
12-10-2013 08:44 AM
Yes, use dummy IPs to test links, but when you install VCS you don't need IPs - it will still work with IPs, but they should be removed to prevent the network being used by anything else.
The link being UP means something is plugged in the NIC, but it is not necessarily plugged into the right place.
Mike
12-10-2013 09:22 AM
Hi Mike,
Tks for the help so far.
I have done what you said and the ping failed.
root # ifconfig -a
lo0: flags=2001000849<UP,LOOPBACK,RUNNING,MULTICAST,IPv4,VIRTUAL> mtu 8232 index 1
inet 127.0.0.1 netmask ff000000
igb0: flags=1000843<UP,BROADCAST,RUNNING,MULTICAST,IPv4> mtu 1500 index 2
inet 10.129.68.47 netmask ffffffc0 broadcast 10.129.68.63
ether 0:10:e0:3e:c3:1c
igb1: flags=1000842<BROADCAST,RUNNING,MULTICAST,IPv4> mtu 1500 index 5
inet 1.1.1.1 netmask ffff0000 broadcast 1.255.255.255
ether 0:10:e0:3e:c3:1d
igb2: flags=1000842<BROADCAST,RUNNING,MULTICAST,IPv4> mtu 1500 index 6
inet 1.1.1.2 netmask ffff0000 broadcast 1.255.255.255
ether 0:10:e0:3e:c3:1e
usbecm0: flags=1000843<UP,BROADCAST,RUNNING,MULTICAST,IPv4> mtu 1500 index 4
inet 169.254.182.77 netmask ffffff00 broadcast 169.254.255.255
ether 2:21:28:57:47:17
root # dladm show-dev
igb0 link: up speed: 1000 Mbps duplex: full
igb1 link: up speed: 1000 Mbps duplex: full
igb2 link: up speed: 1000 Mbps duplex: full
igb3 link: unknown speed: 0 Mbps duplex: unknown
usbecm0 link: up speed: 10 Mbps duplex: full
In bold the interfaces that were unknokn staete before the plubing and these are the ones to be used for the HB.
Have I missed anything?
Used these commands:
root@# ifconfig igb2 plumb
root@ # ifconfig igb2 1.1.1.2
root@ # ifconfig igb2 netmask 255.255.0.0
The ping from the other node failed to the 1.1.1.1 or 1.1.1.2
Tks in advance ,
Joao
12-10-2013 09:28 AM
Hi Joao,
you should give 1.1.1.1 on first node on igb1 & 1.1.1.2 on second node on igb1 ..
similarly 1.1.2.1 on igb2 on first node & 1.1.2.2 on second node on igb2
at this point 1.1.1.1 & 1.1.1.2 are assigned to same server which should ideally ping (on same server) however try above to confirm if connectivity is correct...
G
12-10-2013 09:36 AM
Will this test work if the nodes are connected through using a switch over a VLAN?
12-10-2013 09:40 AM
I cant do this test because the other node the interface has a different name and the cluster is currently running there. So, I cant assign an IP to the interface of the other node and test this...
I can ping on the own machine but this doesnt prove much...
12-10-2013 09:42 AM
If I umplumb it it gets to the unkknown state again.
root@dp-node9 # ifconfig igb1 unplumb
root@dp-node9 # netstat -in
Name Mtu Net/Dest Address Ipkts Ierrs Opkts Oerrs Collis Queue
lo0 8232 127.0.0.0 127.0.0.1 388 0 388 0 0 0
igb0 1500 10.129.68.0 10.129.68.47 15569708 0 7256551 0 0 0
igb2 1500 1.1.0.0 1.1.1.2 12943 0 24 0 0 0
usbecm0 1500 169.254.182.0 169.254.182.77 6 0 1126 0 0 0
root@dp-node9 # dladm show-dev
igb0 link: up speed: 1000 Mbps duplex: full
igb1 link: unknown speed: 1000 Mbps duplex: full
igb2 link: up speed: 1000 Mbps duplex: full
igb3 link: unknown speed: 0 Mbps duplex: unknown
usbecm0 link: up speed: 10 Mbps duplex: full
Is this expected?
12-10-2013 09:43 AM
Hi,
yes, ideally it should work .. because you are configuring the IP address of same subnet to communicate over VLAN ...
if in case above test is not working,
1. recheck connectivity, try connecting via cross cables if possible to eliminate the problem with switch.
2. If cross cable is not available, check the switch logs to confirm if packets are getting dropped.
G
12-10-2013 09:48 AM
Sorry, couldn't see previous comments for some reason ... pinging on self machine is not gonna make much sense ...
Because heartbeat is already up & running in the cluster & if cluster is a production cluster, I wouldn't recommend to try plumbing an IP on heartbeat.
Your ultimate task is to add the node to cluster, so anyways you would need to install VCS packages. You can install VCS packages (download from symantec portal) with that, you will get VRTSllt package.
with VRTSllt installed & all the cluster config files (llttab, llthosts,gabtab) populated, you can then run the dlpiping or lltping tests as explained in first step
G
12-10-2013 09:50 AM
You had 2 issues:
Mike
12-10-2013 09:56 AM
If you can't assign IPs to other node as it is live, then you cant test using IPs. There shouldn't be an issue assigning IPs to a live node from a technical point of view, but running any command on a live server can be risky if you make a mistake (hence why most live systems will have change control)
Mike
12-10-2013 11:27 PM
I agree with Mike - there is no problem adding a IP address to private NICs, even if it is a live cluster.
Plumbing IPs on separate subnets is the quickest and best way to test comms.
12-10-2013 11:39 PM
Tks Marianne. The problem is that I dont have the permission to do it :(
I will install the packages on the new node and use the llt tools and try to figure this out.
Let you know how it goes... Tks for the help so far.
12-22-2013 07:15 PM
Hi,
Was wondering if you've got a chance to test this ?
G
12-23-2013 02:45 AM
Im having problems on getting a license for the new node... As soon as I install the SW in the new node and test I post here.
Tks,
Joao
12-23-2013 08:25 AM
Hi,
I have done some testing here and after make all the configurations: new node is dp-e9
root@dp-e9 # cat /etc/llthosts
0 DP-e5
1 DP-e6
2 DP-e4
3 DP-e8
4 dp-e9
root@dp-e9 # cat /etc/gabtab
/sbin/gabconfig -c -n5
root@dp-e9 # cat /etc/llttab
set-node dp-e9
set-cluster 10000
link igb1 /dev/igb1 - ether - -
link igb2 /dev/igb2 - ether - -
root@dp-node9 # dladm show-dev
igb0 link: up speed: 1000 Mbps duplex: full
igb1 link: up speed: 1000 Mbps duplex: full
igb2 link: up speed: 1000 Mbps duplex: full
igb3 link: unknown speed: 0 Mbps duplex: unknown
usbecm0 link: up speed: 10 Mbps duplex: full
In the other node I have:
root@DP-e6 # /sbin/gabconfig -a
GAB Port Memberships
================================
Port a gen 406917 membership 01234
Port b gen 406915 membership 0123
Port b gen 406915 visible ; 4
Port h gen 406914 membership 0123
Port h gen 406914 visible ; 4
Looks like everyhthing is fine...
Tks for the help.