Network checks fail on 11.0.0.434 to 11.0.1 upgrade

[laserted]

Hi - I'm attempting to upgrade Scalix CE 11.0.0.434 to 11.0.1 now that my webmail users are getting the dreaded "time expired" popup.

The previous edition 11.0.0.434 is runing FINE (the time expired excepted) on RHEL4, with a direct routable IP address on the internet - no routing tables or mapping in the way.

Tarball downloaded, extracted, running installer via CLI (since I'm about 1800 miles away from the server currently). I have not uninstalled anything, I am running the option [1] Upgrade....with all [1..10] component upgrade options (as default) accepted.

Failures:
Network Check Failed - Could Not find a host name in DNS that matches an IP address
Network Check Failed- IP address associated with host name "localhost.localdomain" is 127.0.0.1
Network Check Failed - 127.0.0.1 line contains fully-qualified server hostname (localhost.localdomain) in /etc/hosts, please remove it.

I know Scalix is picky about the hosts file. I am not running local DNS; DNS is in the nameserver of my domain registrar, and does contain the appropriate zone info.

Hosts file contains:
#Obligatory don't touch me note
127.0.0.1 localhost.localdomain localhost
66.93.num.num mail1.mydomain.com

(note that num.num are replaced with real octets....)

hostmane --fqdn returns: mail1.mydomain.com

(Where mydomain is substituted for my actual domain name).

I have made multiple mods in test, restarting the network services afterwards, and the failure is consistent, with the exception if "localhost" is missing, the installer gets more upset....

Thoughts?

Thanks,

Ted.

[kanderson]

what does dig mail.mydomain.com return? Also what does route show?

Kev.

[laserted]

Kev -

; <<>> DiG 9.2.4 <<>> mail1.mydomain.com
;; global options: printcmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 48145
;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 0

;; QUESTION SECTION:
;mail1.mydomain.com. IN A

;; ANSWER SECTION:
mail1.mydomain.com. 297 IN A 66.93.num.num

;; Query time: 17 msec
;; SERVER: 66.92.159.2#53(66.92.159.2)
;; WHEN: Wed Feb 28 19:02:15 2007
;; MSG SIZE rcvd: 52


mydomain replaces my real domain name, and num.num the last two octets, which match the host, netconfig and zone file entry.


Also:
Kernel IP routing table
Destination Gateway Genmask Flags Metric Ref Use Iface
66.93.246.240 * 255.255.255.248 U 0 0 0 eth0
169.254.0.0 * 255.255.0.0 U 0 0 0 eth0
default dsl093-246-241. 0.0.0.0 UG 0 0 0 eth0

Which is as anticipated; the subnet mask is correct as displayed.

Thanks,

Ted.

[kanderson]

Strange, there's no loopback in that route command. What does ifconfig show?

Can you ping 127.0.0.1?

Kev.

[laserted]

# ifconfig
eth0 Link encap:Ethernet HWaddr 00:01:29:92:83:1F
inet addr:66.93.246.245 Bcast:66.93.246.247 Mask:255.255.255.248
inet6 addr: fe80::201:29ff:fe92:831f/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:8977346 errors:0 dropped:0 overruns:0 frame:0
TX packets:10356451 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:1447216491 (1.3 GiB) TX bytes:1664942064 (1.5 GiB)
Interrupt:10 Base address:0x4000

lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
UP LOOPBACK RUNNING MTU:16436 Metric:1
RX packets:2200011 errors:0 dropped:0 overruns:0 frame:0
TX packets:2200011 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:1037105753 (989.0 MiB) TX bytes:1037105753 (989.0 MiB)

# ping 127.0.0.1
PING 127.0.0.1 (127.0.0.1) 56(84) bytes of data.
64 bytes from 127.0.0.1: icmp_seq=0 ttl=64 time=0.062 ms
64 bytes from 127.0.0.1: icmp_seq=1 ttl=64 time=0.078 ms


As noted before, this servcer is in production with 11.0.0.434, and is a standalone node "hot" on the 'net.....

Thoughts?

Ted.

[kanderson]

Can you humor me and type

route add -net 127.0.0.1

and see if that fixes it? You WILL need that loopback route. It will be used by Scalix to talk to sendmail, for example.

Kev.

[laserted]

Um, loopback would already have been configured,, but here goes....

#route add -net 127.0.0.1
STOCADDRT: Invalid Argument
#

To my knowledge, didn't kernels 2.1+ retain the loopback as a dedicated interface instead of a static route? (hence its existence in the ifconfig list)?

Either way, it won't let me add it. (route add lo 127.0.0.1 obviously fails as an "unhonwn host", too....)

Ted

[kanderson]

I'm running a 2.6 kernel, and I see a loopback. Can you ping it?

Kev.

[dougp23]

This might be unrelated, but here's how I do a hosts file for a Scalix install:

pico /etc/hosts (or use vi)
I then delete the 127.0.0.1 line and replace it with
127.0.0.1 {TAB} {TAB} localhost
62.27.27.27 {TAB} FQDN {TAB} Alias

Of course, {TAB} means to press the TAB key.
so my real address line might read
62.62.65.65 {TAB} mymail.mycompany.com {TAB} mymail

I have found this to be the easiest way to prevent the Scalix installer from barfing on the Network check.

Good luck!

[laserted]

Kev - yes, as mentioned above, I can ping 127.0.0.1 , the physical NIC IP, the hostnames, and even externally to the system, so network functionality s fine. All of my access is currently from 1500 miles away on a Putty link.

I don't think it's a problem with the hosts file or an IPTables issue. I have a feeling I am going to stop wasting my time hunting this gremlin down and do a fresh OS and Scalix install, something I can't do remotely, unfortunately.

So...can one save the existing databases and message stores and drop them into the new install? I've seen lots of upgrading from ver10 to 11+, but not from 11.0.0.434 to 11.0.1....most of the install notes don't cover clean sweeps then restoring exiting data....what are your recommendations?

Thanks,

Ted.

[laserted]

Doug - although this isn't my first Scalix install and certainly not my only linux server (I have 8 Centos boxes and 5 Win32 boxes in the server cluster at this particular office), I wiped the contents of the hosts file for this box and gave it a shot - I typically didn't do a double-TAB and have only "localhost" on the local line (I use the single tab, with "localhost.localdomain {tab} localhost" routine....but figured it was worth the effort. The second line of course included my routable net-IP plus the FQDN and alias.

The scalix install came back like a broken record...
Network check failed -- Hosts file is not configured with fully qualified domain name (FQDN)
Network check failed -- Could not find a host name in DNS that matches an IP address
Network check failed -- Could not resolve host name 'localhost.localdomain' in DNS
Network check failed -- 127.0.0.1 line contains server hostname (localhost) in /etc/hosts, please remove it
Dependency check passed.

I'm going to partition one of my test servers for a clean CentOs +Scalix 11.0.1 blank install just to remove any of the abstract possibilities.

Thanks,

Ted.


[quote="dougp23"]This might be unrelated, but here's how I do a hosts file for a Scalix install:

pico /etc/hosts (or use vi)
I then delete the 127.0.0.1 line and replace it with
127.0.0.1 {TAB} {TAB} localhost
62.27.27.27 {TAB} FQDN {TAB} Alias

Of course, {TAB} means to press the TAB key.
so my real address line might read
62.62.65.65 {TAB} mymail.mycompany.com {TAB} mymail

I have found this to be the easiest way to prevent the Scalix installer from barfing on the Network check.

Good luck![/quote]

[kanderson]

Before you do all of that, you might as well try an install of 11.0.2. Also, can you post the last 20 lines or so of /var/log/scalix-installer-<date>.log. There might be something more in there.

If that doesn't work, then yes, you can keep the mailstore. stop scalix with the following commands.

/etc/init.d/scalix stop
/etc/init.d/scalix-postgres stop
/etc/init.d/scalix-tomcat stop

Then tar up /var/opt/scalix, preserving permissions with something like:

tar -zcvpf /backup.tar.gz /var/opt/scalix

Then scp the tarball to somewhere else, test it to be sure you have it all, then go ahead with the OS install. Then scp back and untar and install scalix. You can and should just go straight to 11.0.2, unless you're using external authentication such as Active Directory, or OpenLDAP, in which case stay at 11.0.1 for now.

[ScalixSupport]

Hi!

Just a heads up, make sure you are trying to upgrade Scalix with the package specific to
RHEL4.

Thanks,
Subir

[laserted]

Ok, this took a lot of work, but here's the scoop:
System running on a hot WAN IP address (no hardware routers, bridges, etc), Centos4 with nothing else but Scalix 11. Complete Centos install, including graphical console and all the bells & whistles.
Centos uses the hosts file, plus IPTables, plus it's own Network Setup wizard. The setup wizard has a hosts table as well that doesn't update the /etc/hosts text file......

1) Fresh install Centos4.
2) UnTar Scalix RHEL4 distrib.
3) Change redhat-version contents to reflect "Red Hat Enterprise Linux ES release 4 (NAHANT)" so that the scalix installer will actually run instead of complaining it's an unsupported distro...
4) While the SCalix installer is getting ready to run, change the /etc/hosts file (text version) to read: 67.67.67.67 {tab} mail1.domainname.com {tab} mail1
127.0.0.1 {tab} localhost
(Where 67.....67 is a real, routable IP address, not a Class C LAN or loopback)
Don't forget to SAVE!

5) Add the same 67...67 mail1.domainname.com and mail1 alias settings to the GRAPHICAL Centos Network Setup wizard. IT WILL NOT LET YOU PUT IN A LOOPBACK, so don't bother. The /etc/hosts text file got it. They do not mirror each other!
6) In the same graphical wizard, clear the DNS search suffix, and make sure the hostname is just the short alias (mail1) not the FQDN hostname.
7) Save all this, activate the adapter so it will reastart.
8) Carry on with the Scalix installer.

Whew! That took a lot of painkillers to get through!

At least it's done.

Ted.