What is the name of the first cluster transport adapter?
Hi,
I am new to Solaris and just entering to clusters. pardon if asking any question that is already answered.
I am installing Sun Cluster on 2 X v120 boxes and i have two interfaces eri0 and eri1 on each box.
I gace connected eri0's to a Switch and eri1's using crossover cable.
eri0's are configured for 10.10.7.xxx and eri1's are in 176.16.0.xxx
this was done because I was not able to telnet on network adaptors connected using crossover cable when they were configured for 10.10.7.x.
Now the problem is while installing sun cluster using scinstall, I am asked for Cluster Transport Adaptor name.
The script will not work is I give eri0 or eri1, it says,
Adapter "eri0" is already in use as a public network adapter
Adapter "eri1" is already in use as a public network adapter.
These are the only two adaptors I have and dont know hoe to proceed further.
Now the question are...
1. Do I need to configure virtual adaptors like eri1:1, eri1:2 or is there any trick that I have misssed?
2. What is the common IP for which the cluster will respond once configured? Is it the IP of First Node or do I need to provide any other IP?
Thanks in advance for the help.
Regards,
Ramesh.
[1288 byte] By [
ballu121] at [2007-11-26 9:43:09]
# 1
Ramesh,
You are going to need a minimum of 3 network adapters for Sun Cluster. I don't think you can force it to install with only two without a fair amount of hacking. You'd somehow have to convince the installer that you have a network adapter that is not working (may be linking an eri2 device to eri1 temporaily might work - I've never tried it though)
If you did have 3 adapters, you would only configure a network address on one public network per node. The other two would be plumbed in by Sun Cluster during the install.
Tim
# 2
Can I work with virtual IP'S?Say, I will configure only one adaptor for public IP (eri0)and provide two ip's on eti1 (eri1, eri1:0)
# 3
No, Sun Cluster would have created all of its required virtual IPs on the adapters for you. All you need to have set up is the public network.Tim
# 4
Hi,
I tried using eri1 and eri1:1 as transport adaptors for cluster. after I unplumb eri1 only eri1 was accepted and not eri1:1. So I am assuming that virtual interfaces cannot be used.
My sun fire v120 box is having two Optical Fiber Cable Connectors.. Can I use them for cluster interconnection instead of network interface eri0 or eri1?
If yes, Please guide me.
# 5
Sorry, but you can't do this unless you have a minimum of 3 physical adapters. (or are prepared to do a lot of hacking of the install scripts).In short, this just isn't going to work.Tim
# 6
Hi Ramesh,I am also facing the same problem, but can I kill the process at this time and it will not disturb further configuration of the cluster? I will add one more interface after killing the scinstall and shutdown of the server.RegardsAmit Bist
Ammi at 2007-7-7 0:43:29 >
# 7
Killing and restarting scinstall later will not cause a problem.Tim
# 8
Hi TimHow does cluster configuration, scinstall, detects quorum drive?RegardsAmit Bist
Ammi at 2007-7-7 0:43:29 >
# 9
Hi,Anyone replied to you about the last question about How does cluster configuration, scinstall, detects quorum drive?Sxx
SXX at 2007-7-7 0:43:29 >
# 10
Ignore my previous msg. Now, how do you solve the shared storage issue in VMWare. I am using free VMWare Server 1.05, not ESX VMWare.Thanks.
SXX at 2007-7-7 0:43:29 >
# 11
I doubt if you will get this to work. I've heard from a colleague that the basic product has issues. He managed to get SC working with ESX though.(expect a blog posting soon)Tim
# 12
I doubt if the vmware server supports the scsi reservations.
So you are always safe to use a single node cluster, than you can ignore the shared storage problem. Just add virtual disks and you are set.
BTW to get rid of a syslog overrun with pm_tick delay you should modify your kernel.notice to kernel.warning.
AS long as you have a 64 bit vm it will work this way in a single node cluster.
Detlef
# 13
Hi, Tim and Detlef,
There is no issue to setup/config SCSI HD on VMWare workstation 4.5 onwards and VM server 1.0X.
I guess the issue is how do we setup/config a "shared storage" under Solaris. In Windows, we can easily achieve it by modifying *.vmx config file to make a hard disk visiable by both servers. So,
(1) Unless we can do the same as Windows *.vmx solution; or
(2) We need to make iSCSI work so that we can use free open source software, "openfiler" for Solaris acting as a NAS/SAN device.
A complete cluster with proper shared storage (hence, proper qorum device.) will be very nice. We can even use this infrastructure to do more serious POCs.
Regards.
SXX at 2007-7-7 0:43:29 >
# 14
Hi,
the problem starts with the special call so scsi reservations, not with the scsi disk.
vmware server talks in their documents how to make a bus virtual tu support scsi reservations, no chance with the worstations.
You may have to "disable" failfast panics by hashing out
/usr/cluster/lib/sc/cmm_ctl -d -f in the cluster start methods.
I am not specific here by intent, because then you are vulnerable to split brain, amnesia, ...
This is totally unsupported and may work, but making mistakes will corrupt your data. If it works fine, treat it with extreme care.
For the quorum problems the quorum server helps, so You then have to start an other vm on your machine to be the quorum server.
For demonstration purpose I am personally using single node clusters and found them sufficient.
I have heard, that the description in the vmware server guide enables scsi2 reservations only, so you are stuck to 2 node clusters, or you have to change the reservation type which is only doable in SC3.2.
Kind Regards
Detlef
# 15
The reason I talked about vm SCSI hd is that I wanted to build a SAN like shared storage both for quorum and data. I need a full-fledged clustering infra to access the impact of data/msg after MQ clustering fail-over. With on node cluster, you only demo how you install/setup/config Sun cluster.
Also, for you to have one vm cluster, we wouldn't even bother to have scsi, right?
So, my question again: how you relate your scsi reservation post to the actual SAN like shared storage issue? Let's assume we don't have physical SAN device.
I am working on making iSCSI work in Solaris 10 x64 + openfiler. Hope someone post this as well.
thanks,
SXXa at 2007-7-21 15:19:25 >
# 16
Somehow I couldn't see the msg I replied to Detlef, so here I summerized it again:
(1) We don't have scsi shared storage hardware in place, so it doesn't help even I make it work;
(2) I hope I can make iSCSI work so that I can use openfiler (software version of SAN like shared storage device that supports iSCSI.)
Hope anyone can post this related blog.
Tks. Sam
SXXa at 2007-7-21 15:19:26 >
# 17
Hi,
The problem with the shared storage is severe in the vmware owrkstation, you might have quite good chance in vmware server, but I have not tried it right yet. A collegue of mine tried it with the caveats in my latest post. However the most likely road to succcess is Vmware ESX here, because ESX offers a feature which RDM (raw device mappings).
If you just want to test a failover of the MQ clustering in a single node cluster, you can failover the MQ between two full root zones. This should be sufficient for a test.
Detlef
