QuantumATK Forum

QuantumATK => Installation and License Questions => Topic started by: ams_nanolab on September 20, 2013, 07:37

Title: Installation problem in 13.8.b1
Post by: ams_nanolab on September 20, 2013, 07:37
Dear Sir,
We are trying to install ATK 13.8.b1 on our RHEL 6.3 64bit version. The GUI is saying that the installation is complete but when we try to run vnl it is giving the following error (attached)
Title: Re: Installation problem in 13.8.b1
Post by: Anders Blom on September 20, 2013, 11:46
Try "which vnl" - is it really trying to start Virtual NanoLab...? It doesn't quite make sense that it attempts to run VNL in parallel over MPI...???

Otherwise and anyway, try to run it not as root, and make sure the path is updated to reflect the new version.
Title: Re: Installation problem in 13.8.b1
Post by: ams_nanolab on September 20, 2013, 12:49
which vnl ... is showing the correct path , we tried to run not as root as well... the same error keeps coming up.  ???
Title: Re: Installation problem in 13.8.b1
Post by: ams_nanolab on September 20, 2013, 13:19
We uninstalled 13.8.b1 and installed 12.8.2 it works just fine. I think there's a bug in your installer, kindly correct the issue.
Title: Re: Installation problem in 13.8.b1
Post by: Anders Blom on September 20, 2013, 14:33
The only time before I saw this issue was discussed and solved here:
http://quantumwise.com/forum/index.php?topic=1360.0#.UjxAED-mWRA
But it involved running "mpiexec", I really see no reason why running purely "vnl" should have this effect.

I'm speculating if you have an issue in /etc/hosts (or similar) which means you can't map the hostname "nano6" to its IP number, or something like that...
Title: Re: Installation problem in 13.8.b1
Post by: ams_nanolab on September 20, 2013, 19:50
Ok, will try this... let u know if it works.  ;D
Title: Re: Installation problem in 13.8.b1
Post by: Dipankar Saha on September 22, 2013, 22:33
A. Blom >
Sir, as u speculated...the problem may be in /etc/hosts....., there I found... it is mentioned two localhost.local domain names... say local domain 4 and 6... for local domain 4 address is perfectly fine...127.0.0.1 , but for the other local domain (local domain 6) ..the address is a junk ( : :1  or something like that) .... Now could it be, by any chance, the reason behind the problem (as mentioned by ams_nanolab) that we are facing ???
Title: Re: Installation problem in 13.8.b1
Post by: Anders Blom on September 22, 2013, 23:20
No, I don't think so, the other address is the IPv6 address for the machine, and the first one is for IPv4.

However, the main question is, if "nano6" is present in the hosts table?
Title: Re: Installation problem in 13.8.b1
Post by: Dipankar Saha on September 22, 2013, 23:34
Sir, can u please tell me....where should I check... to verify ....whether 'nano6' is there in the host table or not???
Title: Re: Installation problem in 13.8.b1
Post by: Anders Blom on September 22, 2013, 23:38
I meant in /etc/hosts. Test if you can "ping nano6".
Title: Re: Installation problem in 13.8.b1
Post by: Dipankar Saha on September 22, 2013, 23:50
Yes, I can ping...when I mention the IP-add of nano6....it is pinging....
Title: Re: Installation problem in 13.8.b1
Post by: Anders Blom on September 23, 2013, 00:02
Actually, I meant the command

ping nano6

(not by IP address).

But I'm anyway grasping at straws, I really can't explain the behavior since I don't see why MPI is even called at startup of "vnl"... We did upgrade the MPICH2 library in ATK for this release, but on the other hand it's running fine on a lot of other machines, so still it has to be something related to your setup that is special.
Title: Re: Installation problem in 13.8.b1
Post by: Dipankar Saha on September 23, 2013, 00:14
May be.... :) /
B d way ...."ping nano6" showing "unknown host" ..../ But how can dis be the issue ? (may be I am not getting it)...Specially when I tried installing ATK 12.8.2...on the very same machine...with out any alternation in network setup...it ran perfectly fine....!!!/ Now, while I'm trying to install 13.8.b1 .... then, running 'vnl' becomes the problematic..

Anyways.... :)
Title: Re: Installation problem in 13.8.b1
Post by: Anders Blom on September 23, 2013, 08:35
I suppose this doesn't have much to do with ATK in itself, but let's hope things work better if you add nano6 to the /etc/hosts file:

1.2.3.4    nano6

As mentioned, we upgraded MPICH2 in ATK between 12.8 and 13.8, which apparently triggers this behavior. Thanks for reporting it, it was quite unexpected. It's not an error, but good to know in case someone else has similar issues.
Title: Re: Installation problem in 13.8.b1
Post by: Dipankar Saha on September 23, 2013, 14:30
Yes....this was the solution... All I needed to do...is to add the corresponding IP add along with 'nano6' in the /etc/hosts .....  :) Thank you sir (A. Blom) for helping me out.... :)
Title: Re: Installation problem in 13.8.b1
Post by: Anders Blom on September 23, 2013, 14:40
I'm very happy about that. For the benefit of other users let me just state the error message you observed in text here, to make it searchable:

Quote
Fatal error in MPI_Init: Other MPI error, error stack:
MPIR_Init_thread(392)..............:
MPID_Init(139).....................: channel initialization failed
MPIDI_CH3_Init(38).................:
MPID_nem_init(234).................:
MPID_nem_tcp_init(108).............:
MPID_nem_tcp_get_business_card(346):
MPID_nem_tcp_init(305).............: gethostbyname failed, nano6 (errno 1)

And, again, in case you see this while running mpiexec, besides checking /etc/hosts, also check your machine file (cf. http://quantumwise.com/forum/index.php?topic=1360.0#.UkA1XIbdeKx)
Title: Re: Installation problem in 13.8.b1
Post by: ams_nanolab on September 23, 2013, 19:35
thanks Dr.Blom.