Installataion within a vserver

[moe]

Hello,

Iam trying to install the Scalix Community Edition 9.4.2.4 within a vserver, but the installer wont let me go on. It fails while Network Check, the log-message is

Network check failed -- Msg: Could not determine network interface configuration for eth0

Starting in debug-mode shows this:

Scalix Installer - starting version 9.4.2.7...
INFO:merlin:Starting Scalix Installer (Community Edition)
DEBUG:root:Startup property dictionary {'adv': True, 'test_mode': False, 'no_syschecks': False, 'exec': './scalix-installer-community', 'no_root': False, 'rpm_dir': '/root/scalix-9.4.2-community-intel-rh/software/scalix_server', 'log_dir': '/var/log', 'start_screen_id': -1, 'debug': True, 'error_on_syscheck_fail': True, 'cwd': '/root/scalix-9.4.2-community-intel-rh'}
Scalix Installer (Community Edition) - audit log file is /var/log/scalix-installer-20060124.log
INFO:merlin:Current locale is and default is
DEBUG:root:Loading screens
DEBUG:root:==> Entering screen WizardWelcome ============
DEBUG:root:WizardWelcome:debug mode True
DEBUG:root:==> Leaving screen WizardWelcome
DEBUG:root:==> Entering screen WizardInstallationType ============
DEBUG:root:WizardInstallationType:debug mode True
INFO:merlin:Saving installation type: installall mode: typical
INFO:merlin:Operation passed: OS release check Fedora Core release 4 (Stentz)
INFO:merlin:Operation passed: Memory check 776808
INFO:merlin:Operation passed: Disk space check 1811216
DEBUG:root:FunctionThread calling determine_packages with args (<sxdata.Data instance at 0xb6f8522c>,)
INFO:merlin:Searching /root/scalix-9.4.2-community-intel-rh/software/scalix_server for relevant Scalix packages
WARNING:merlin:File scalix-websplats-configurator-py22-9.4.2.5-1.noarch.rpm is not a Scalix package
DEBUG:root:Looking for minimum version for component scalix-websplats-configurator-py22 and couldn't find it
INFO:merlin:File scalix-sac-9.4.2.5-1.noarch.rpm is a Scalix file, for this platform and valid version
INFO:merlin:File scalix-swa-9.4.0.52-1.noarch.rpm is a Scalix file, for this platform and valid version
WARNING:merlin:File scalix-server-community-9.4.2.4-0.1.rhel4.i386.rpm is not for this platform
WARNING:merlin:File scalix-websplats-configurator-py23-9.4.2.5-1.noarch.rpm is not a Scalix package
DEBUG:root:Looking for minimum version for component scalix-websplats-configurator-py23 and couldn't find it
WARNING:merlin:File scalix-websplats-configurator-py24-9.4.2.5-1.noarch.rpm is not a Scalix package
DEBUG:root:Looking for minimum version for component scalix-websplats-configurator-py24 and couldn't find it
DEBUG:root:Packages in dir (sorted): [['scalix-swa', ['9.4.0.52', '4524069', 'noarch', 'linux', '1', '(none)', 'scalix-swa-9.4.0.52-1.noarch.rpm']], ['scalix-server-community', ['9.4.2.4', '71069212', 'i386', 'linux', '0.1.fc4', 'FC4', 'scalix-server-community-9.4.2.4-0.1.fc4.i386.rpm']], ['scalix-sac', ['9.4.2.5', '11053478', 'noarch', 'linux', '1', '(none)', 'scalix-sac-9.4.2.5-1.noarch.rpm']]]
DEBUG:root:Latest packages in dir: {'scalix-sac': ['9.4.2.5', '11053478', 'noarch', 'linux', '1', '(none)', 'scalix-sac-9.4.2.5-1.noarch.rpm'], 'scalix-swa': ['9.4.0.52', '4524069', 'noarch', 'linux', '1', '(none)', 'scalix-swa-9.4.0.52-1.noarch.rpm'], 'scalix-server-community': ['9.4.2.4', '71069212', 'i386', 'linux', '0.1.fc4', 'FC4', 'scalix-server-community-9.4.2.4-0.1.fc4.i386.rpm']}
DEBUG:root:Not upgradeable: {}
INFO:merlin:Tomcat was running when we started: False
DEBUG:root:==> Leaving screen WizardInstallationType
DEBUG:root:==> Entering screen WizardAllComponents ============
DEBUG:root:WizardAllComponents:debug mode True
DEBUG:root:WizardAllComponents:Installable: {'scalix-sac': ['9.4.2.5', '11053478', 'noarch', 'linux', '1', '(none)', 'scalix-sac-9.4.2.5-1.noarch.rpm'], 'scalix-swa': ['9.4.0.52', '4524069', 'noarch', 'linux', '1', '(none)', 'scalix-swa-9.4.0.52-1.noarch.rpm'], 'scalix-server-community': ['9.4.2.4', '71069212', 'i386', 'linux', '0.1.fc4', 'FC4', 'scalix-server-community-9.4.2.4-0.1.fc4.i386.rpm']}
WARNING:merlin:Required components not found: []
INFO:merlin:Components selected to install: {'scalix-sac': ['9.4.2.5', '11053478', 'noarch', 'linux', '1', '(none)', 'scalix-sac-9.4.2.5-1.noarch.rpm'], 'scalix-swa': ['9.4.0.52', '4524069', 'noarch', 'linux', '1', '(none)', 'scalix-swa-9.4.0.52-1.noarch.rpm'], 'scalix-server-community': ['9.4.2.4', '71069212', 'i386', 'linux', '0.1.fc4', 'FC4', 'scalix-server-community-9.4.2.4-0.1.fc4.i386.rpm']}
INFO:merlin:Components to upgrade: {}
DEBUG:root:==> Leaving screen WizardAllComponents
DEBUG:root:==> Entering screen WizardSystemCheck ============
DEBUG:root:WizardSystemCheck:debug mode True
DEBUG:root:WizardSystemCheck:We've checked already False and changed component selection False
INFO:merlin:System check start
DEBUG:root:This host has the following IP addresses: ['10.6.1.236']
DEBUG:root:This machine's host name 'scalix.my.domain' and IP address '10.6.1.236' is set in DNS correctly
INFO:merlin:Network interface is actually eth0
DEBUG:root:Checking DHCP for interface eth0
ERROR:merlin:Operation failed: Network check failed -- Msg: Could not determine network interface configuration for eth0
DEBUG:root:Server is not installed so using default scalix path
DEBUG:root:Message store checking: /var/opt/scalix/sys/version exists: False
INFO:merlin:Operation passed: Filesystem check passed.
WARNING:merlin:JAVA_HOME is set but the directory does not exist
DEBUG:root:Java version at /usr/java/j2sdk1.4.2_10 is 1.4.2_10
INFO:merlin:Java and Tomcat check info -- JAVA_HOME environment variable not set but guessed at /usr/java/j2sdk1.4.2_10
INFO:merlin:Java and Tomcat check info -- Java version 1.4.2_10
INFO:merlin:Java and Tomcat check info -- Default Tomcat directory is None
WARNING:merlin:Operation warning: Java and Tomcat check warning -- No valid Tomcat directory found. You may install Tomcat now or use the Installer in the Configuration steps.
DEBUG:root:Component list for mode install {'scalix-sac': ['9.4.2.5', '11053478', 'noarch', 'linux', '1', '(none)', 'scalix-sac-9.4.2.5-1.noarch.rpm'], 'scalix-swa': ['9.4.0.52', '4524069', 'noarch', 'linux', '1', '(none)', 'scalix-swa-9.4.0.52-1.noarch.rpm'], 'scalix-server-community': ['9.4.2.4', '71069212', 'i386', 'linux', '0.1.fc4', 'FC4', 'scalix-server-community-9.4.2.4-0.1.fc4.i386.rpm']}
INFO:merlin:Apache appears to be installed.
INFO:merlin:Operation passed: Dependency check passed.
INFO:merlin:Operation passed: Running services check passed.
INFO:merlin:System check end
INFO:merlin:System check result: fail

I am sure the installer doesnt understand the network configuration, in the vserver the interface eth0 is unconfigured the "real" interface is eth0:<vservername> an alias of eth0.

The debug-output shows a variable 'error_on_syscheck_fail': True, how can I influence this variable?
While searching in this forum Ive found somethin about a cli-only-install, but this isnt available anymore in this release?

Greetings and excuse my english ;-) Maurice

P.S. Iam the first one who's trying to install scalix in a vserver or why I havent found anything on google?

[ScalixSupport]

Hi Maurice,

It is indeed possible to install and run Scalix in a virtual machine. It is not, however, a recommended solution for a number of reasons; performance and reliability to name two.

What product are you using for your virtual machine? The product I use allows me to configure eth0 as if it was a stand alone server. To resolve this you will have to have eth0 configured in a standard Linux way. Is the vm on the network in such a way as you can ping other machines, browse the web etc?

What is the output of ifconfig?

No, you can't influence the variable. The network check has to pass.

Release 10 will have a cli but it too requires that the network check passes.

Regards,
Don

[moe]

Hi Don, thanks for the fast answer.

This Installation of scalix is for testing purposes, in a productive environment is a dedicated server of course the better solution.

I'm running vserver on a Gentoo Host, the network setup is on host-side also done with ethernet aliases. Other virtual machines using tun/tab or similar solutions and there is the network-device a "real" eth0 on client-side.. But this machine ist in generally for testing purposes of variuos things and software, and vserver is the simplest and ressource effective solution.

The output of ifconfig is:

Code: Select all

eth0      Link encap:Ethernet  HWaddr 00:50:BA:6D:4B:06 
          UP BROADCAST NOTRAILERS RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:91634 errors:0 dropped:0 overruns:0 frame:0
          TX packets:107610 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:36267740 (34.5 MiB)  TX bytes:68994243 (65.7 MiB)
          Interrupt:11 Base address:0xc000

eth0:sca  Link encap:Ethernet  HWaddr 00:50:BA:6D:4B:06 
          inet addr:10.6.1.236  Bcast:0.0.0.0  Mask:255.255.255.255
          UP BROADCAST NOTRAILERS RUNNING MULTICAST  MTU:1500  Metric:1
          Interrupt:11 Base address:0xc000


The network-setup is completely done on host-side, but is fully functional as if it would be a standalone server. I can ping to other machines, and other machines can also ping the vm and connect to running servers a.s.o.
The only thing is, there is no lo-device in the guest, and that the guest cannot control the interface, setting another address or options with ifconfig has to be done on the host.

Ive tried to fake the output of ifconfig, with replacing the original executable with a shell-script which outputs a "normal" ifconfig-output, but this has no effect.

Greetings Maurice

[ScalixSupport]

Hi Maurice,

I'm afraid I am not familiar with Gentoo. Vmware allows you to configure a nic as if there was a real device present. Again I will reiterate what I said yesterday. To resolve this you will have to have eth0 configured in a standard Linux way. We are looking for inet addr in the results of ifconfig for eth0.

Best wishes,
Don

[moe]

Hi again,

Hmm I feel the installer is a little bit restrictive. Why i am not allowed to install if my ethernet-device has not the name eth0? This is not only a problem in my special case, other virtualizing-environments also often have other names for emulated net devices. In OpenVZ a.k.a. Virtuozzo its venet0 for example. VMWare is a very good program, but its an emulator of a whole computer, not a virtualizer, so it needs much more ressources than other technologies in this application. Also a server could have more than one ethernet-device and eth0 is used for other services or even unused. Another case could be a VPN-Connection so that the net-device is tun0, a.s.o.
Beside of this, most of people who installing such a service like scalix are administrators, who know what they are doing, why there is'nt the possibility to go on even in case of failures? Of course you want to have all requirements satisfied, to eliminate support-questions which are caused by wrong system configurations. But on the other hand nearly everyone can bypass the restrictive installer with installing on another computer and transferring the installation to the "bad" computer (or do some other tweaking/disassembling). Of course its harder, and avoids the installation on systems which dont fit the requirements in most cases, but it also avoid some installations on system which are fully capable of running scalix, but unfortunately the installer doesnt knows about.
I don't want to raise a quarrel or critizing a product which I don't know, I only want to show why I believe this installer is too restrictive in this point. Iam working as administrator and consultant, we are working heavily with vservers, to test various software (complete linux distributions, or even services like scalix) and often we prepare an installation of a complete server in a vserver an put this later on real hardware of the customer. And maybe I am the first which have had this problem with the installation, but I'm surely not the last one. Virtualizing is much practical and a growing trend in the it-industry, see the hype about xen or an interesting blog about for examples. Xen wouldnt have this problem with a "wrong named" network device, but there are so many possibilities, and the installer can't know about anything. A good adminstrator (should) knows his system better than the installer, but he will be needlessly restrained.

Regards, Maurice

[ScalixSupport]

Hi Maurice,

I hear what you are saying and you make many valid points. One of them caused quite a chuckle here in support.

Beside of this, most of people who installing such a service like scalix are administrators, who know what they are doing...

If you have a look at the postings on this forum you will find that, by and large, the opposite is true. This is not to denigrate those that are posting. In fact many of them have been quite forthcoming in admitting their newness to Linux. As such we have worked very hard to make the installer as straightforward as possible obviating possible support issues. In fact the release 10 installer due out soon will ask very few questions and does it's best to surmount many of the issues people have encountered.

To the issue at hand. I will repeat myself and say what you are trying to attempt is not a "supported" configuration. I'll try to do what I can to help you overcome this issue but there may come a time when I have to throw up my hands.

Before we go any further the vserver is one of our supported Linux distributions correct?

My first question is trying to understand your vserver more fully. Thanks for the link to the vserver documentation. Sorry, I don't have the time to read that right now. What I'd like to know is why you can't ifconfig eth0 and bridge it to the host?

If that is absolutely not possible we may be able to get past this hurdle. The network check determines the default interface and in this case is recognizing eth0 not the virtual eth0:sac. Once it sees the default is eth0 the installer reads /etc/sysconfig/network-scripts/ifcfg-eth0. Try creating this file and inputing the configuration for eth0:sac. You may have to edit /etc/sysconfig/networking/devices/ifcfg-eth0 and
/etc/sysconfig/networking/profiles/default/ifcfg-eth0 too. I'm not sure.

Give that a go and let me know the results. I've got my fingers crossed.

Regards,
Don

[mephisto]

Hi Maurice,

have you made any progress with you effords? I'm planning to do something quite similar and vserver seems to be less of a bloat than vmware.

Regards,

Mephisto

[mephisto]

With OpenVZ you can move a physical network adapter to the virtual private server, so if you have multiple NICs in your host, you can e.g. make eth1 accessible only on the vps:

Code: Select all

vzctl set 101 --netdev_add eth1 --save

@Don: is the installer limited to eth0 or does it probe all "normal" ethX interfaces?

[mephisto]

Seems I just answered my own question: The installer (9.2.2 eval) does not complain even if I run it on OpenVZ with only a venet0:0 interface.

[ScalixSupport]

Way to go mephisto! Thanks for that.

Cheers,
Don

[SiD3WiNDR]

I've just installed Scalix 10 in a VServer running Fedora Core. After fixing up the hosts file like the installer wanted, the network check was actually the biggest hurdle. The missing link was indeed creating the ifcfg-eth0 file mentioned above, which is missing on standard vserver installation because the networking is setup by the host system rather than via the guest system's scripts.

Why doesn't the installer simply parse "ifconfig eth0"'s output or rather "ip a"'s output, but actually goes and read the system's config files? That's not a real way to determine if the address is assigned to the system, is it...

I set onboot=no, as it won't work to run it anyway.

PS, only on the host system the interface is seen as "eth0:scalix"; inside the vserver i just have a nice normal regular "eth0". :)

[Renegade]

Hiho, i'm working on the same problem.

I'm using openSuSE 10.1 guest on a debian etch host.

with the ifcfg-eth0 workaround it worked smoothly, but i have some problems with the localhost. Because it doesn't exist!

Cause of that, i get the following error:

Code: Select all

Mar 26 15:58:28 mail sendmail[12804]: l2PDwNlQ002928: SYSERR(root): ldap_init/ldap_bind failed to localhost in map ldapsx: Unknown error 325


Code: Select all

mail:~ # ifconfig
eth0      Link encap:Ethernet  HWaddr 00:02:A5:1A:2C:7B
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:266254 errors:0 dropped:0 overruns:0 frame:0
          TX packets:90601 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:35696965 (34.0 Mb)  TX bytes:44651898 (42.5 Mb)

eth0:v02  Link encap:Ethernet  HWaddr 00:02:A5:1A:2C:7B
          inet addr:89.149.199.191  Bcast:89.149.199.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1

mail:~ # hostname
mail.renegadesworld.de

mail:~ # hostname -s
mail

mail:~ # hostname -f
mail.renegadesworld.de

mail:~ # cat /etc/hosts
#
# hosts         This file describes a number of hostname-to-address
#               mappings for the TCP/IP subsystem.  It is mostly
#               used at boot time, when no name servers are running.
#               On small systems, this file can be used instead of a
#               "named" name server.
# Syntax:
#
# IP-Address  Full-Qualified-Hostname  Short-Hostname
#

127.0.0.1       localhost
127.0.0.2       mail.renegadesworld.de mail
89.149.199.191  mail.renegadesworld.de mail

mail:~ # uname -a
Linux mail.renegadesworld.de 2.6.18-3-vserver-686 #1 SMP Mon Dec 4 18:50:30 UTC 2006 i686 i686 i386 GNU/Linux


[mephisto]

Your installation doesn't have the loopback interface (lo) activated. You need to set START_LOOPBACK="yes" in rc.config - or do that with yast - check your manual for details.

[Renegade]

I know that there is no loopback device, THAT IS THE PROBLEM. Think about it, i'm using a vserver!

How can i get the loopback device within the vserver?