PDA

View Full Version : SLES 11 SP2 for VMWare drop odbc connections



jesussanmol
08-Nov-2012, 17:06
Hi all,

Sorry for my english.

I've installed Sles11 SP2 for VMWare on my vmware 5.0U1 farm enterprise plus and I'm having problems with external clients (VPN) that connects to informix database via ODBC.

The behaviour its very extrange because at first, users can connect and work without problems but at not determinated moment drops connection and its impossible connect again (no ping, etc...). SuseFirewall2 its stopped and no rules are added but I noticed that If I start and stop susefirewall services starts to work again but in minutes drops again.

I have no idea that whats its happens, with Sles10 on physical machines works perfectly.

I'm very worried with that because its our database production server.

I apreciatte any help

Thank you in advance

ab
08-Nov-2012, 17:54
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

So... SLES works, but VMware environments have a problem? What is
VMware doing with your connections?

The easiest way to see the firewall settings is to get the output of the
following.... it should be pretty short if you have the firewall "stopped":

sudo /usr/sbin/iptables -nvL

Also, even a default firewall configuration doesn't cut of connections
intermittently, and any blocking it does shows up in /var/log/firewall
so as a result I'm pretty sure this isn't SLES.

Good luck.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://www.enigmail.net/

iQIcBAEBAgAGBQJQm+OvAAoJEF+XTK08PnB5k9QP/2byD3GNryDJZ2wBVW9cypyy
p4rxvwTA5Qj59T+xhOIwXXxMXNJGdfXp7aZz4lEKz6mH0lOOCy FTc3f0kTux7jBP
wK6sy+dnaBEVNNlpgtnXt1MZ4OkIFFqqVIgngHgnkGZjW5zKby +uhDkYkIB8IrKk
s+DmHg2keuCeG6YpAD2hp29oFudK5BcjR7frR3FjmJr6r+xlLM hp1FUtOL7/xgve
Ka9sEW1P6ln37Xc7EozX5aPqhiObB+0yJEAc9JKD4KWu5JH87C lr3fqBZOUWvHbY
/YVA9ES0R/qqrLiVJZCM0ICrCq3VWXlc+NWoDrWxtOy7e97ixUjFCIL2D/kPvuq7
TZfLOTBuBY+z+/S8XE3sMOgR2EpYEYxW4ldbIwCFzvmBwCa6TJhm1/Qi5XWD8ylg
gLnTvlMX0Apk9pOaJpTOn+H/cE9mTLXqY8/W4IhJyitYQKYuGQExO/KG+SXL8MhX
jDA+9O43crz9gAnf3sUzsJxFzqbsGf+jG1XAZ79+WvOzIQoXdz 4CBvitbDyPauSB
vkpy9gNG4OPM1mt0Jexdg+27/8xvVAx+4LEd41lTZcKI7KhopkZ+8KecKoIFR3jo
Ru4S4mVl9RD3gv8O4RYVN5+ZMrejFKcvhNkoTOhtagGknkviKc dvnHTcHHGzbDT/
IE4dSvct+7CzRXK/DM+3
=nKUv
-----END PGP SIGNATURE-----

cjcox
08-Nov-2012, 23:22
This "could" be relevant.. but the guest side offloading often times conflicts
with the server/hypervisor side NIC offloading driver... so disable offloading
on the client side.

SUSE 11+ Linux Variants (disabling on boot)

To disable the options as a part of the boot up process. Edit
/etc/sysconfig/network/ifcfg-eth<num> (where <num> is the identified for the
virtual NIC you are targeting). Inside of that file add/edit the variable
ETHTOOL_OPTIONS. For example, if the target virtual NIC is eth0, edit
/etc/sysconfig/network/ifcfg-eth0 and make sure it has the following variable set:

ETHTOOL_OPTIONS='-K eth0 gso off tso off'

To do this live you can just issue the options above directly to ethtool for you
interface (above assumed eth0).

We have 10Gbit NICs in our hosts and without disabling the offloading our
performance degrades to Kbits (yes.. that bad)

Again.. not saying this is the source of your headache.. but what I listed here
is relevant for Gbit NICs as well as 10Gbit.

I am also running ESX 5.0U1


On 11/08/2012 10:14 AM, jesussanmol wrote:
>
> Hi all,
>
> Sorry for my english.
>
> I've installed Sles11 SP2 for VMWare on my vmware 5.0U1 farm enterprise
> plus and I'm having problems with external clients (VPN) that connects
> to informix database via ODBC.
>
> The behaviour its very extrange because at first, users can connect and
> work without problems but at not determinated moment drops connection
> and its impossible connect again (no ping, etc...). SuseFirewall2 its
> stopped and no rules are added but I noticed that If I start and stop
> susefirewall services starts to work again but in minutes drops again.
>
> I have no idea that whats its happens, with Sles10 on physical machines
> works perfectly.
>
> I'm very worried with that because its our database production server.
>
> I apreciatte any help
>
> Thank you in advance
>
>

jesussanmol
09-Nov-2012, 09:25
Thank you for your anser, I've checked that firewall its off there no rules active in iptables. I'm going to try TSO off.

I was thinking if kernel net.ipv4.* security parameters could be the cause but i'm not sure, its very difficult try in productions enviroments.

Thank you very much for you help

jesussanmol
12-Nov-2012, 09:45
Hi all, any solution doesn't work.

I'll try everything without success, sles still dropping informix odbc connections from external networks. I don't know what to do.

Thanks

jmozdzen
12-Nov-2012, 13:11
Hi jesussanmol,

the symptoms sound like a case I had debugged a few days ago - does your setup include a separate VPN router that does run parallel to the default router, from SLES11SP2's point of view?

Regards,
Jens

jesussanmol
16-Nov-2012, 09:22
Hi jmozdzen,

Yes I have a separate VPN Router from default router, its that the problem? I'm trying to understand but I dont know what to do for fix that situation. I'm thinking to install sles 10 sp2 but i'm not sure to do that and everything works fine.

Its there anu solution from Novell about that?

Regards

jmozdzen
16-Nov-2012, 10:38
Hi jesussanmol,

> Yes I have a separate VPN Router from default router, its that the problem?

my suspicion is towards trouble with ICMP redirect handling in the Linux kernel that is shipped with SLES11SP2 - if your VPN router is in the same network as your SLES11SP2 server and you have no explicit route set for the VPN network, you're likely to run into the problems I'm thinking of. (see this recent thread: https://forums.suse.com/showthread.php?1903-SLES-11-SP2-for-VMware-lose-connectivity)

*If* I'm right, the cure is simple: At the SLES11SP2 server, add an explicit route for the VPN network, pointing at the VPN router.

Of course, as I had just debugged such a case, this may be a case of selective recognition... like when you have a new car, you seem to notice that the same model is everywhere around you ;)

Regards,
Jens

forresn
17-Nov-2012, 03:04
Hi jesussanmol,

> Yes I have a separate VPN Router from default router, its that the problem?

my suspicion is towards trouble with ICMP redirect handling in the Linux kernel that is shipped with SLES11SP2 - if your VPN router is in the same network as your SLES11SP2 server and you have no explicit route set for the VPN network, you're likely to run into the problems I'm thinking of. (see this recent thread: https://forums.suse.com/showthread.php?1903-SLES-11-SP2-for-VMware-lose-connectivity)

*If* I'm right, the cure is simple: At the SLES11SP2 server, add an explicit route for the VPN network, pointing at the VPN router.

Of course, as I had just debugged such a case, this may be a case of selective recognition... like when you have a new car, you seem to notice that the same model is everywhere around you ;)

Regards,
Jens

I think we are having similar issues since upgrading to OES 11 SP1/SLES11SP2 19 days ago. The problem really didn't start seeing itself until earlier this week. I was getting reports of computers not able to reach the server. I had them ping the server and get no response, but the computer right next to it would get a response. Changing the IP address of the computer worked though! A tcpdump of the server showed the computer that was not working requesting a ping, but no response was sent. I'll look into the explicit route and see if that's an issue. Could we expect a fix for this in the future?

jesussanmol
19-Nov-2012, 12:16
I hope a fix will be release, if don't i have to migrate to other distribution to fix this situation.

jmozdzen
20-Nov-2012, 18:52
Hi,

did you try setting the explicit route and it didn't work?

Regards,
Jens

jesussanmol
21-Nov-2012, 16:08
Hi, Yes I set explicit routes and the did not work. I cant understand why drop connections; for example clients in the same subnet are working normal but in some time a few client works and others doesn't, the I restart SuseFirewall service and the stop again (because Susefirewall is stopped) and everything works again for all clients. I can't find a solution.

KBOYLE
21-Nov-2012, 17:28
jesussanmol wrote:

> I have no idea that whats its happens, with Sles10 on physical
> machines works perfectly.

Take a look at this TID.

TID 7007649 Applying SLES 11 SP 1 Causing Communication Issues
http://www.novell.com/support/kb/doc.php?id=7007649

Your issue may not be a physical/virtual problem but may be due to the
way routing has changed after SLES 11.

--
Kevin Boyle - Knowledge Partner
If you find this post helpful and are using the web interface,
show your appreciation and click on the star below...

jmozdzen
27-Nov-2012, 17:07
Hi,

I find that TID worded confusingly:


The max value from conf/{all,interface}/rp_filter is used
when doing source validation on the {interface}. (Which means it can be set

on each individual interface, or globally. The global setting will override the

individual interface settings).

"max value from conf/{all,interface}/rp_filter" is something totally different from "The global setting will override the
individual interface settings" as setting "2" at the interface level will not be overridden by "1" at the global level?!?

Regards,
Jens