[Users] Shutdown problems

[HostGIS Support]

I emailed on the topic before, and have never found a solution -- nor 
indeed, more than one other corroboration of the problem's existence. 
But now, I have freed up a while server with OpenVZ where we can 
experiment with it at will.


The problem: Shutting down a VPS gives me a timeout after several 
minutes. Although all processes in the container are dead, the container 
itself will not finish shutting down. The veth device never goes down, 
the container cannot be restarted, the phantom VPS will hang around 
until I power-cycle the server. This interrupts shutdowns too: init 0 
and reboot never, ever work; they do nothing, they don't turn anything 
off; and I have to pull the plug.


Worse, this happens reliably -- I don't dare shut down a VPS unless it's 
a migration, and I can manually complete the migration and startup, then 
power-cycle the origin HN.


BUT... Now we have a machine and some IPs with OpenVZ, and my current 
project is to figure this thing out so we can reboot with confidence. 
Where do we start and who's with me? :)


--
HostGIS, Open Source solutions for the global GIS community
Greg Allensworth - SysAdmin, Programmer, GIS Person, Security
Network+   Server+   A+   Security+

No one cares if you can back up — only if you can recover.

___
Users mailing list
Users@openvz.org
https://openvz.org/mailman/listinfo/users

Re: [Users] Shutdown problems

[HostGIS Support]

To clarify further on versions:

The HN is 2.6.24 ovz009.1 on Fedora 9.

We must use 2.6.24 despite its development status because 2.6.18 lacks 
support for AMCC/3ware RAID controllers. Aside from this shutdown issue, 
we have used it for 14 months now under high loads without issue. Aside 
from this shutdown issue, we consider it stable and production-grade.


--
HostGIS, Open Source solutions for the global GIS community
Greg Allensworth - SysAdmin, Programmer, GIS Person, Security
Network+   Server+   A+   Security+

No one cares if you can back up — only if you can recover.

___
Users mailing list
Users@openvz.org
https://openvz.org/mailman/listinfo/users

Re: [Users] Shutdown problems

[HostGIS Support]

have you searched for a bug report or checked the forums?  
Has this issue been reported?


I'll take the if you don't have anything nice to say, don't say 
anything clause for my previous experience with forums. Maybe the 
OpenVZ forum is unlike any other, but honestly the thought of posting to 
a forum didn't occur to me with any degree of seriousness. I did post to 
this list several months ago, and got no replies.


But, I am surprised that the bug I filed last year is not visible. Boy 
do I feel like a luser; I bet I closed the wrong tab or something and 
never finished filing it -- no wonder nobody replied. Per Thorsten's 
request I will indeed file the bug report; maybe the kernel folks can 
verify that it's been solved or maybe something really weird is going on.




Red Hat updates their 2.6.18 kernel [...]

 in the process they backport a lot of drivers
 [...] current OpenVZ RHEL5-based kernel to see if it
 is compatible with your hardware?

According to the kernel config file inside the RPM, RHEL5-2.6.18 now has 
the 3W-9000 driver. Good to know; that may indeed be a realistic option 
and an easy fix.


So, can you please clarify: The RHEL5 kernel is usable and recommended 
if I'm NOT running RHEL5?




You said you are using Fedora 9 and that was EOLed a while ago.


Yeah, and it was the latest thing 14 months ago when we deployed this 
thing. I can't say I'm completely pleased with that fast a retirement 
cycle. You say that CentOS is a fine base OS? I have worked with it in a 
few occasions, and find it similar enough to Fedora that I like it.


Changing out base OSs isn't something I'd do lightly, but is something I 
would consider as a longer-term plan if it improved long-term support or 
fixed this bug we're having. Beyond this one bug, things are going great.




You mentioned you are using veth devices


Correct. In my initial setup of the pilot systems veth was the only 
thing that worked, so we went with it. Why do you ask, and what do you 
mean about it being a security issue? (I did ask about ARP and IP 
spoofing on the list some months back, and that one also got no replies.)




What distro or distros are you running in your containers?


This seems to happen equally with both of our offerings currently 
deployed: Ubuntu 8.04 and HostGIS Linux 4.2 (for your purposes, think 
Slamd64 12).


--
HostGIS, Open Source solutions for the global GIS community
Greg Allensworth - SysAdmin, Programmer, GIS Person, Security
Network+   Server+   A+   Security+

No one cares if you can back up — only if you can recover.

___
Users mailing list
Users@openvz.org
https://openvz.org/mailman/listinfo/users