Re: [Vserver] jakarta-tomcat doesn't work with vserver?

2004-05-20 Thread Chris Wilson
Hi Markus,

 i have some strange problem.
 I'm just installing jakarta-tomcat-5.0.24
 and j2sdk1.4.2_04 in a vserver.
 When i want to start tomcat the following error occours:
[...]
 Error occurred during initialization of VM
 Could not reserve enough space for code cache

I can report that it definitely works with Tomcat 4.1.30 on stock 2.4.25 
with VS 1.27 patch:

[EMAIL PROTECTED] root]# uname -a
Linux converge..xxx 2.4.25-vs1.27 #1 SMP Sat Apr 10 09:31:38 BST 
2004 i686 i686 i386 GNU/Linux

[EMAIL PROTECTED] root]# java -version
java version 1.4.2_04
Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_04-b05)
Java HotSpot(TM) Client VM (build 1.4.2_04-b05, mixed mode)

[EMAIL PROTECTED] root]# ps auxww | grep tomcat | head -1
tomcat8960  0.0 16.2 243400 146844 ? SMay07   0:04 
/usr/java/j2sdk1.4.2_04/bin/java 
-Djava.endorsed.dirs=/home/tomcat/common/endorsed -classpath 
/usr/java/j2sdk1.4.2_04/lib/tools.jar:/home/tomcat/bin/bootstrap.jar 
-Dcatalina.base=/home/tomcat -Dcatalina.home=/home/tomcat 
-Djava.io.tmpdir=/home/tomcat/temp org.apache.catalina.startup.Bootstrap 
start

So you may want to try a stock kernel and/or Tomcat 4.1.30?

Cheers, Chris.
-- 
_  __ __ _
 / __/ / ,__(_)_  | Chris Wilson -- UNIX Firewall Lead Developer |
/ (_  ,\/ _/ /_ \ | NetServers.co.uk http://www.netservers.co.uk |
\__/_/_/_//_/___/ | 21 Signet Court, Cambridge, UK. 01223 576516 |
___
Vserver mailing list
[EMAIL PROTECTED]
http://list.linux-vserver.org/mailman/listinfo/vserver


Re: [Vserver] Experimental Version

2004-04-20 Thread Chris Wilson
Hi Herbert,

  other stuff can be found here:
  http://vserver.13thfloor.at/Experimental/
 
 yes, actually it's vs1.9.0pre10.3 ... ;)

Well, actually it's vs1.9.0pre11 right now, but let's not split hairs :-)

 with vs1.9.0pre10* you can actually disable the
 proc security from the menuconfig (or *config)

Hmm, it's a shame that that's all you can do, because at least when using 
the stable tools, I can't stop a vserver with this feature enabled:

[EMAIL PROTECTED] vservers]# vserver distcc stop
Stopping the virtual server distcc
Error: /proc must be mounted and readable
  To mount /proc at boot you need an /etc/fstab line like:
  /proc   /proc   procdefaults
  In the meantime, `mount /proc /proc -t proc'
  To set the permissions, `chmod 755 /proc'
Server distcc is not running

Thanks to you and Bjoern for all your help! I'm working on getting things 
up and running with 2.6.6rc1 and the pre11 patch now (and /proc security 
disabled).

Cheers, Chris.
-- 
_  __ __ _
 / __/ / ,__(_)_  | Chris Wilson -- UNIX Firewall Lead Developer |
/ (_  ,\/ _/ /_ \ | NetServers.co.uk http://www.netservers.co.uk |
\__/_/_/_//_/___/ | 21 Signet Court, Cambridge, UK. 01223 576516 |

___
Vserver mailing list
[EMAIL PROTECTED]
http://list.linux-vserver.org/mailman/listinfo/vserver


Re: [Vserver] Experimental Version

2004-04-20 Thread Chris Wilson
Hi Herbert,

 with vs1.9.0pre10* you can actually disable the
 proc security from the menuconfig (or *config)

OK, I built a 2.6.6-rc1 kernel with the -pre11 patch, and it Oopses when I 
try to enter a vserver:

[EMAIL PROTECTED] vservers]# vserver distcc1 enter
/sbin/ifconfig eth1:distcc1 192.168.3.181 netmask 255.255.255.0 broadcast 
192.168.3.255
SIOCSIFADDR: File exists
SIOCSIFFLAGS: Cannot assign requested address
SIOCSIFNETMASK: Cannot assign requested address
SIOCSIFBRDADDR: Cannot assign requested address
SIOCSIFFLAGS: Cannot assign requested address
ipv4root is now 192.168.3.181
Host name is now distcc1.netservers.co.uk
New security context is 49155
Segmentation fault

linux1 kernel: kernel BUG at include/linux/vinline.h:62!
linux1 kernel: invalid operand:  [#3]
linux1 kernel: PREEMPT
linux1 kernel: CPU:0
linux1 kernel: EIP:0060:[c0113385]Not tainted
linux1 kernel: EFLAGS: 00010286   (2.6.6-rc1-vs1.9.0pre11)
linux1 kernel: EIP is at mm_init+0xe2/0x101
linux1 kernel: eax: e610bc00   ebx:    ecx: edfeff80   edx: 
e610bc00
linux1 kernel: esi: e4b25ea4   edi: dfd8a76c   ebp: dfd8a580   esp: 
e4b25dfc
linux1 kernel: ds: 007b   es: 007b   ss: 0068
linux1 kernel: Process save_s_context (pid: 2082, threadinfo=e4b24000 
task=defdc330)
linux1 kernel: Stack: dfd8a580 e4b25e6c 0001  dfd8a580 
  dfd8a6e0
linux1 kernel:dfd8a6e0     
  
linux1 kernel:     
  
linux1 kernel: Call Trace:
linux1 kernel:  [c01136ea] copy_mm+0xe7/0x427
linux1 kernel:  [c0114321] copy_process+0x453/0xb7c
linux1 kernel:  [c01593ea] do_pipe+0x185/0x205
linux1 kernel:  [c0114a9a] do_fork+0x50/0x16d
linux1 kernel:  [c01b398a] copy_to_user+0x3e/0x4e
linux1 kernel:  [c014c9ba] sys_llseek+0x9f/0xc4
linux1 kernel:  [c0104387] sys_clone+0x41/0x45
linux1 kernel:  [c010575d] sysenter_past_esp+0x52/0x71
linux1 kernel:
linux1 kernel: Code: 0f 0b 3e 00 5f e6 28 c0 eb dc a1 ec 1f 35 c0 89 6c 
24 04 89

Any ideas?

Cheers, Chris.
-- 
_  __ __ _
 / __/ / ,__(_)_  | Chris Wilson -- UNIX Firewall Lead Developer |
/ (_  ,\/ _/ /_ \ | NetServers.co.uk http://www.netservers.co.uk |
\__/_/_/_//_/___/ | 21 Signet Court, Cambridge, UK. 01223 576516 |

___
Vserver mailing list
[EMAIL PROTECTED]
http://list.linux-vserver.org/mailman/listinfo/vserver


Re: [Vserver] Lost access to vserver IPs

2004-04-15 Thread Chris Wilson
Hi Cathy,

  are you 100% sure that this isn't some hidden
  arp table flushing / router not routing issue?
 
 I had similar problems (similarly affecting only the vserver IPs, not the 
 master server) that coincided with a router upgrade in the data center.  I 
 second Herbert's suggestion to check the router. :)  HTH.

I will, but I definitely observed repeated incoming ARP requests for an IP
that I thought was configured on the server, although I neglected to
capture evidence. The addresses had not previously been assigned, so I
don't think there could be an ARP table entry to flush, and having 
restarted the vservers it once again became possible to access them 
without any changes to the firewall.

I will investigate further when I can afford potential downtime, 
and capture a tcpdump and the output of ip link show, ip route show 
and arp -n to prove that the machine is not responding to ARPs. Is there 
anything else I should capture?

Is it possible that vserver X stop does not always take down the right 
interface?

Cheers, Chris.
-- 
_  __ __ _
 / __/ / ,__(_)_  | Chris Wilson -- UNIX Firewall Lead Developer |
/ (_  ,\/ _/ /_ \ | NetServers.co.uk http://www.netservers.co.uk |
\__/_/_/_//_/___/ | 21 Signet Court, Cambridge, UK. 01223 576516 |

___
Vserver mailing list
[EMAIL PROTECTED]
http://list.linux-vserver.org/mailman/listinfo/vserver


Re: [Vserver] Lost access to vserver IPs

2004-04-15 Thread Chris Wilson
Hi Thomas,

 Herbert helped me to trace down the problem: We are running
 the master server and the vservers in different ip subnets:

Yes, that is _exactly_ what we are doing.

 (ip from the iproute2 suite)  you can figure out which of the ip alias
 interfaces are secondary. These ip aliases will be removed if you
 shut down the corresponding primary interface.

I did check, and noticed the difference, but I wasn't sure about the 
significance of it, except to try to rearrange the order of addresses to 
make a different one master.

 we had some problems with nasty side effects when stopping
 one specific vserver: all other vservers on the same master lost
 their network connectivity.

Well, we didn't lose all of them, but quite a few.

 The kernel ip stack treats the first ip address within a scope
 as primary and deletes all secondary ip addresses within
 this scope when the primary address is taken down.

Ouch, that is a very nasty thing for it to do. Maybe we should be using 
ip addr add and ip addr del to add/remove addresses instead of 
ifconfig'ing interfaces up and down? Is this (mis)feature related to 
backwards compatibility with the old way of doing aliases?

 (ifconfig doesn't show the primary/secondary feature, but ip from
 iproute2 does)

Yes, we have only a few primary addresses, and many secondaries.

I have to discuss it with my boss, but your solution looks interesting,
although non-standard. If this really is the problem, then two other
solutions pose themselves.

Change the main IP address of the master server to be in the same subnet
as the vservers (not scalable, what happens when we run out of addresses
in this subnet?)

Add a VLAN interface for each vserver, to isolate them (and fix the
problems with getting the wrong netmask due to the scripts' assumptions
that the master is in the same subnet as the vservers). Does this also 
make CAP_NET_RAW somewhat safer? (are you still restricted to the same 
physical interface that your address is bound to, or can you spoof/listen 
to any packets on any interface that you want?)

Thanks very much for your help,

Cheers, Chris.
-- 
_  __ __ _
 / __/ / ,__(_)_  | Chris Wilson -- UNIX Firewall Lead Developer |
/ (_  ,\/ _/ /_ \ | NetServers.co.uk http://www.netservers.co.uk |
\__/_/_/_//_/___/ | 21 Signet Court, Cambridge, UK. 01223 576516 |

___
Vserver mailing list
[EMAIL PROTECTED]
http://list.linux-vserver.org/mailman/listinfo/vserver


Re: [Vserver] Lost access to vserver IPs

2004-04-15 Thread Chris Wilson
Hi Herbert,

  Is it possible that vserver X stop does not always take down the right 
  interface?
 
 yes, under yet unknown circumstances, it happens
 that the vserver tools (or the ip/ifconfig utility)
 creates the alias as primary IP, in which case a
 vserver down (for this alias) will take down the
 entire interface (which is the right thing to do
 for primary ip addrs, which should not be used by
 vserver tools) ...

Yes, that is exactly what happened. Thomas was exactly right with his 
explanation, which is that when the vservers are in a different subnet to 
the main server, the first vserver to come up gets a non-secondary 
address, with the results that you describe.

 this can cause other servers to become disfunctional
 after a vserver stop (which you didn't mention ;)

Well, I did say that I restarted a vserver, I assumed that a vserver stop 
was implicit in that :-)

 you can verify such a mis-configuration with
  ip addr show on the host, which should only list secondary aliases for
 the vservers IPs ...

I did, I saw that vservers did indeed have primary addresses, and I fixed
it by moving the master server into the same subnet as all the vservers,
restarting networking, and then entering each vserver to get it to
recreate its addresses (I also had to stop the vservers that already had
primary addresses, before restarting networking, to stop them from keeping
their addresses across the restart).

Thanks guys! I think that's solved the problem.

Cheers, Chris.
-- 
_  __ __ _
 / __/ / ,__(_)_  | Chris Wilson -- UNIX Firewall Lead Developer |
/ (_  ,\/ _/ /_ \ | NetServers.co.uk http://www.netservers.co.uk |
\__/_/_/_//_/___/ | 21 Signet Court, Cambridge, UK. 01223 576516 |


___
Vserver mailing list
[EMAIL PROTECTED]
http://list.linux-vserver.org/mailman/listinfo/vserver