Home » Mailing lists » Users » Debian: recommended kernel
| Debian: recommended kernel [message #46208] |
Wed, 02 May 2012 16:31  |
Roman Haefeli
Messages: 9
Registered: May 2012
|
Junior Member |
|
|
Hi all
We're running OpenVZ on Debian Squeeze with the kernel shipped by
Debian.
Several sources recommend to use RHEL 6 stable kernel. Is it recommended
to use it also on Debian stable? If so, how should it be installed? The
wiki has links to rpm files only, it seems.
The reason I ask is that we're considering switching kernel, since the
Debian OpenVZ kernel seems to have issues. Specifically, we suffer from
the same problem as described here:
http://lists.debian.org/debian-kernel/2011/12/msg00689.html
Cheers
Roman
|
|
|
|
|
|
| Re: Debian: recommended kernel [message #46210 is a reply to message #46209] |
Wed, 02 May 2012 19:18  |
coolcold
Messages: 11
Registered: May 2007
Location: Moscow, Russia
|
Junior Member |
|
|
On Wed, May 2, 2012 at 9:39 PM, Timh B <timh@shiwebs.net> wrote:
> This was linked earlier this week;
> https://github.com/CoolCold/tools/blob/master/openvz/kernel/ create-ovz-kernel-for-debian.sh
>
> Might be useful for you if you wish to get a debianized openvz-kernel.
>
> If you really want to run debian I would suggest looking into proxmox
> instead since they build the stable rpm-kernels for debian.
May be good, choice, yes. Not tried myself, but reading their maillist.
>
> Good luck!
>
> //T
>
> On Wed, May 2, 2012 18:31, Roman Haefeli wrote:
>> Hi all
>>
>> We're running OpenVZ on Debian Squeeze with the kernel shipped by
>> Debian.
>>
>> Several sources recommend to use RHEL 6 stable kernel. Is it recommended
>> to use it also on Debian stable? If so, how should it be installed? The
>> wiki has links to rpm files only, it seems.
>>
>> The reason I ask is that we're considering switching kernel, since the
>> Debian OpenVZ kernel seems to have issues. Specifically, we suffer from
>> the same problem as described here:
>> http://lists.debian.org/debian-kernel/2011/12/msg00689.html
I was forced to switch into RHEL based kernels instead of Debian ones
because of another bug -
http://bugzilla.openvz.org/show_bug.cgi?id=1747 - I had connection
resets while on tcp sockets.
>>
>> Cheers
>> Roman
>>
>>
> -- Timh
>
>
--
Best regards,
[COOLCOLD-RIPN]
|
|
|
|
| Re: Debian: recommended kernel [message #46211 is a reply to message #46209] |
Wed, 02 May 2012 19:19 |
jjs - mainphrame
Messages: 44
Registered: January 2012
|
Member |
|
|
I second the proxmox idea - it is an excellent platform, I recently
downloaded it to take a look and was quite impressed. It's a more
comprehensive solution than openvz+ovz-web-panel. I'm sold on it and will
deploy it in future. It does require x86_64 hardware though.
Joe
On Wed, May 2, 2012 at 10:39 AM, Timh B <timh@shiwebs.net> wrote:
> This was linked earlier this week;
>
> https://github.com/CoolCold/tools/blob/master/openvz/kernel/ create-ovz-kernel-for-debian.sh
>
> Might be useful for you if you wish to get a debianized openvz-kernel.
>
> If you really want to run debian I would suggest looking into proxmox
> instead since they build the stable rpm-kernels for debian.
>
> Good luck!
>
> //T
>
> On Wed, May 2, 2012 18:31, Roman Haefeli wrote:
> > Hi all
> >
> > We're running OpenVZ on Debian Squeeze with the kernel shipped by
> > Debian.
> >
> > Several sources recommend to use RHEL 6 stable kernel. Is it recommended
> > to use it also on Debian stable? If so, how should it be installed? The
> > wiki has links to rpm files only, it seems.
> >
> > The reason I ask is that we're considering switching kernel, since the
> > Debian OpenVZ kernel seems to have issues. Specifically, we suffer from
> > the same problem as described here:
> > http://lists.debian.org/debian-kernel/2011/12/msg00689.html
> >
> > Cheers
> > Roman
> >
> >
> -- Timh
>
>
|
|
|
|
|
|
| Re: Debian: recommended kernel [message #46216 is a reply to message #46215] |
Thu, 03 May 2012 06:32 |
coolcold
Messages: 11
Registered: May 2007
Location: Moscow, Russia
|
Junior Member |
|
|
On Thu, May 3, 2012 at 9:14 AM, Kir Kolyshkin <kir@openvz.org> wrote:
> On 05/02/2012 09:39 PM, Timh B wrote:
>>
>> This was linked earlier this week;
>>
>> https://github.com/CoolCold/tools/blob/master/openvz/kernel/ create-ovz-kernel-for-debian.sh
>>
>> Might be useful for you if you wish to get a debianized openvz-kernel.
>
>
> Yet another solution is to use alien:
> http://wiki.openvz.org/Install_kernel_from_RPM_on_Debian_6.0
Well, alien works for binary part of kernel, but when I've tried to
install kernel headers that way, i had files conflicts, that's why
that script above was born.
>
> The bad thing about it is it's not a "native" way. The good thing is you
> have the same bit-by-bit binary kernel which we test a lot.
>
>
>>
>> If you really want to run debian I would suggest looking into proxmox
>> instead since they build the stable rpm-kernels for debian.
>>
>> Good luck!
>>
>> //T
>>
>> On Wed, May 2, 2012 18:31, Roman Haefeli wrote:
>>>
>>> Hi all
>>>
>>> We're running OpenVZ on Debian Squeeze with the kernel shipped by
>>> Debian.
>>>
>>> Several sources recommend to use RHEL 6 stable kernel. Is it recommended
>>> to use it also on Debian stable? If so, how should it be installed? The
>>> wiki has links to rpm files only, it seems.
>>>
>>> The reason I ask is that we're considering switching kernel, since the
>>> Debian OpenVZ kernel seems to have issues. Specifically, we suffer from
>>> the same problem as described here:
>>> http://lists.debian.org/debian-kernel/2011/12/msg00689.html
>>>
>>> Cheers
>>> Roman
>>>
>
--
Best regards,
[COOLCOLD-RIPN]
|
|
|
|
| Re: Debian: recommended kernel [message #46226 is a reply to message #46215] |
Fri, 04 May 2012 09:58 |
Roman Haefeli
Messages: 9
Registered: May 2012
|
Junior Member |
|
|
Thanks for all the responses!
On Thu, 2012-05-03 at 09:14 +0400, Kir Kolyshkin wrote:
> On 05/02/2012 09:39 PM, Timh B wrote:
> > This was linked earlier this week;
> > https://github.com/CoolCold/tools/blob/master/openvz/kernel/ create-ovz-kernel-for-debian.sh
> >
> > Might be useful for you if you wish to get a debianized openvz-kernel.
>
> Yet another solution is to use alien:
> http://wiki.openvz.org/Install_kernel_from_RPM_on_Debian_6.0
>
> The bad thing about it is it's not a "native" way. The good thing is you
> have the same bit-by-bit binary kernel which we test a lot.
In the meanwhile I have tried two of the suggested solutions:
* RHEL6-kernel (debianized with alien)
* proxmox-kernel from their repository (which for the 2.6.32 version
says it is based on RHEL6)
(I haven't looked yet into building my own with CoolCold's script)
I didn't have troubles to install either of those. However, both expose
the exact same problem. The network bridges vzbr0 and vzbr1 do not work
correctly and I need those for the veth devices of the containers.
Networking without those bridges seem to work fine, but when the bridges
are active, TCP connections fail to work. UDP seems to be not affected,
both inbound and outbound connections work. However, with TCP I can only
establish outbound connections, but as soon as the connection is
established, nothing is received on the other end anymore. Nor can I
send from outside to the machine with vzbr0.
When I switch back to Debian's OpenVZ kernel, the problems disappear,
but come back, when I boot into the RHEL6 or proxmox kernel. I wonder if
really the kernel is the problem or rather some network/bridge
configuration. I assume others do run bridges successfully with the
aforementioned kernels also run CTs with veths successfully. Is that
correct?
Roman
This is my network configuration on said machine:
eth2 eth3 physical NICs
| |
+--+--+
|
bond0
|
+---+-----+----------+
| | |
vlan20 vlan30 vlan31 VLANs
| |
vzbr0 vzbr1 bridges
DMZ local(NFS)
---
# The loopback network interface
auto lo
iface lo inet loopback
# Fasst die beiden physischen 10Gb-Ports zu
# einem Interface zusammen
auto bond0
iface bond0 inet manual
bond-mode 4
bond-miimon 100
bond_xmit_hash_policy layer2+3
bond_lacp_rate slow
slaves eth2 eth3
# VLAN für DMZ (über bond0)
auto vlan20
iface vlan20 inet manual
vlan_raw_device bond0
address 0.0.0.0
# VLAN für NFS Netz (über bond0)
auto vlan30
iface vlan30 inet manual
vlan_raw_device bond0
address 0.0.0.0
# VLAN für iSCSI Netz (über bond0)
auto vlan31
iface vlan31 inet static
vlan_raw_device bond0
address 10.10.11.92
network 10.10.11.0
netmask 255.255.255.0
# Bridge für VZ (DMZ)
auto vzbr0
iface vzbr0 inet static
bridge_ports vlan20
bridge_maxwait 2
address 195.176.254.92
network 195.176.254.0
netmask 255.255.255.128
broadcast 195.176.254.127
gateway 195.176.254.1
# Bridge für VZ (NFS Netz)
auto vzbr1
iface vzbr1 inet static
bridge_ports vlan30
bridge_maxwait 2
address 10.10.10.92
network 10.10.10.0
netmask 255.255.255.0
---
|
|
|
|
| Re: Debian: recommended kernel [message #46281 is a reply to message #46226] |
Wed, 09 May 2012 07:33 |
Roman Haefeli
Messages: 9
Registered: May 2012
|
Junior Member |
|
|
On Fri, 2012-05-04 at 11:58 +0200, Roman Haefeli wrote:
> Thanks for all the responses!
>
> On Thu, 2012-05-03 at 09:14 +0400, Kir Kolyshkin wrote:
> > On 05/02/2012 09:39 PM, Timh B wrote:
> > > This was linked earlier this week;
> > > https://github.com/CoolCold/tools/blob/master/openvz/kernel/ create-ovz-kernel-for-debian.sh
> > >
> > > Might be useful for you if you wish to get a debianized openvz-kernel.
> >
> > Yet another solution is to use alien:
> > http://wiki.openvz.org/Install_kernel_from_RPM_on_Debian_6.0
> >
> > The bad thing about it is it's not a "native" way. The good thing is you
> > have the same bit-by-bit binary kernel which we test a lot.
>
> In the meanwhile I have tried two of the suggested solutions:
>
> * RHEL6-kernel (debianized with alien)
> * proxmox-kernel from their repository (which for the 2.6.32 version
> says it is based on RHEL6)
>
> (I haven't looked yet into building my own with CoolCold's script)
>
> I didn't have troubles to install either of those. However, both expose
> the exact same problem. The network bridges vzbr0 and vzbr1 do not work
> correctly and I need those for the veth devices of the containers.
> Networking without those bridges seem to work fine, but when the bridges
> are active, TCP connections fail to work. UDP seems to be not affected,
> both inbound and outbound connections work. However, with TCP I can only
> establish outbound connections, but as soon as the connection is
> established, nothing is received on the other end anymore. Nor can I
> send from outside to the machine with vzbr0.
Our problem looks very similar to the one described here:
https://bugzilla.redhat.com/show_bug.cgi?id=623199
Could it be that current stable OpenVZ suffers from the same issue? Or
asked differently: Is the current stable OpenVZ kernel based on RHEL 6.0
which does not contain the fixes as described in:
http://rhn.redhat.com/errata/RHSA-2011-0542.html
Roman
|
|
|
|
| Re: Debian: recommended kernel [message #46282 is a reply to message #46208] |
Wed, 09 May 2012 07:49 |
Roman Haefeli
Messages: 9
Registered: May 2012
|
Junior Member |
|
|
On Wed, 2012-05-02 at 18:31 +0200, Roman Haefeli wrote:
> Hi all
>
> We're running OpenVZ on Debian Squeeze with the kernel shipped by
> Debian.
>
> Several sources recommend to use RHEL 6 stable kernel. Is it recommended
> to use it also on Debian stable? If so, how should it be installed? The
> wiki has links to rpm files only, it seems.
>
> The reason I ask is that we're considering switching kernel, since the
> Debian OpenVZ kernel seems to have issues. Specifically, we suffer from
> the same problem as described here:
> http://lists.debian.org/debian-kernel/2011/12/msg00689.html
It turns out that RHEL 6 based OpenVZ kernel causes more severe issues
than the one from Debian (see my last mail in this thread).
Actually, we found a simple work-around for our initial problem (see
above link): We turn off locking with the mount.nfs option 'nolock'.
We have a HA cluster with corosync/pacemaker running, which manages the
CTs which are running on an NFS export shared across all nodes. As the
HA layer makes sure to not run duplicate instances of CTs, there seems
to be no reason to have NFS locking enabled.
Roman
|
|
|
|
|
|
|
|
|
|
| Re: Debian: recommended kernel [message #46289 is a reply to message #46281] |
Thu, 10 May 2012 16:06 |
|
|
On 05/09/2012 11:33 AM, Roman Haefeli wrote:
> On Fri, 2012-05-04 at 11:58 +0200, Roman Haefeli wrote:
>> Thanks for all the responses!
>>
>> On Thu, 2012-05-03 at 09:14 +0400, Kir Kolyshkin wrote:
>>> On 05/02/2012 09:39 PM, Timh B wrote:
>>>> This was linked earlier this week;
>>>> https://github.com/CoolCold/tools/blob/master/openvz/kernel/ create-ovz-kernel-for-debian.sh
>>>>
>>>> Might be useful for you if you wish to get a debianized openvz-kernel.
>>> Yet another solution is to use alien:
>>> http://wiki.openvz.org/Install_kernel_from_RPM_on_Debian_6.0
>>>
>>> The bad thing about it is it's not a "native" way. The good thing is you
>>> have the same bit-by-bit binary kernel which we test a lot.
>> In the meanwhile I have tried two of the suggested solutions:
>>
>> * RHEL6-kernel (debianized with alien)
>> * proxmox-kernel from their repository (which for the 2.6.32 version
>> says it is based on RHEL6)
>>
>> (I haven't looked yet into building my own with CoolCold's script)
>>
>> I didn't have troubles to install either of those. However, both expose
>> the exact same problem. The network bridges vzbr0 and vzbr1 do not work
>> correctly and I need those for the veth devices of the containers.
>> Networking without those bridges seem to work fine, but when the bridges
>> are active, TCP connections fail to work. UDP seems to be not affected,
>> both inbound and outbound connections work. However, with TCP I can only
>> establish outbound connections, but as soon as the connection is
>> established, nothing is received on the other end anymore. Nor can I
>> send from outside to the machine with vzbr0.
> Our problem looks very similar to the one described here:
> https://bugzilla.redhat.com/show_bug.cgi?id=623199
>
> Could it be that current stable OpenVZ suffers from the same issue? Or
> asked differently: Is the current stable OpenVZ kernel based on RHEL 6.0
> which does not contain the fixes as described in:
> http://rhn.redhat.com/errata/RHSA-2011-0542.html
If you go to http://wiki.openvz.org and put RHSA 2011 0542 into a search
box you will get an answer,
which is kernel 042stab013.1 released a bit less than a year ago, end of
May 2011.
So I am pretty sure the bug you are referring to is fixed.
Anyway, please file a new bug to bugzilla.openvz.org, and provide a
simple scenario of how to reproduce it (please write a complete report).
Kir Kolyshkin
|
|
|
|
Goto Forum:
Current Time: Tue May 07 19:55:19 GMT 2024
Total time taken to generate the page: 0.01511 seconds
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
