PVE1.9 Kernel 2.6.32-6 and Proxmox Mailgateway = Fail

[rengiared]

hi
i think i'm the victim of some sort of bug with the new kernel-version 2.6.32-6.

if u use the free proxmox mailgateway version in a kvm-host with pve kernel version 2.6.32-6 the host hangs up on start.
some times the boot is successful but after a few minutes the mailgateway hangs up with 100% cpu-usage in pve
the moment i switched back to 2.6.32-4 everything is good and works as usually

unfortunately i can't test more cause its a productive server but maybe i have time next week to test it with a test-environment to find out more.
as soon as i have more info, i'm gonna post it here

keep up the great work!

 

[tom]

as you can imagine we test our own mail gateway on pve, so there are no known issues with this combination. pls post information about your hardware, raid controller, ...

what kernel do you run in the mail gateway?

 

[rengiared]

it's a Dell PowerEdge R310
Xeon X3440, 8GB RAM, 2x250GB SATA 7k2rpm, SAS 6iR - Raid 1
3x VMs running on it
pve-manager/1.9/6542
atm Linux 2.6.32-4-pve #1 SMP Mon May 9 12:59:57 CEST 2011

proxmox-mailgateway/2.6/4540
Linux 2.6.27-2-proxmox
as it comes out of the box (a 3-domain-license is in the budget-plan for next year)

i got another possiblity in my head how i can test it. i do a test-restore on my 2 other PVE Server (Dell R710 und R715) who are running with 1.9 and 2.6.32-6 nicely i'm going to post more as soon as i have more information

btw: if javascript is inactive the breaks in the text are deleted (in the forum)

 

[rengiared]

short update: the restore on another machine showed no problem, with the difference that the mailgateway was not productive and not in the same DMZ as normal
i'm going to try the -6 kernel again on the productive machine but it will need a few days until i have the chance und the time to do this late night

 

[rengiared]

today i had the chance to test it again with 2.6.32-6 and unfortunately it failed again
after a few minutes the mailgateway stops working and won't react to anything

on the host it looks like this


now i'm back on 2.6.32-4 again, and also on my two other machines cause it seems like 2.6.32-6 has also a problem with 2 or more v-cpus and windows server 2008 r2 sp1 (always gettin a bluescreen with 0x00000101 Clock Watchdog Error)
maybe both errors have the same origin? (cause the mailgateway has also assigned 2 v-cpu's)

 

[tom]

As also you windows box gots issues I think you got a problem on you host, not with the mail gateway. the mail gateway works, also win2088r2sp1 has no known issues.
can you test the latest kernel from the pvetest repo?

 

[rengiared]

the mailgateway host is a Dell R310, the host of the 2008r2sp1 servers is a r715
the bluescreen of the 2k8 server says "A clock interrupt was not received on a secondary processor within the allocated time interval."

and yes sure, i can test it in the next days on the R310 with the mailgateways
can you explane it how i get the pvetest repo kernel and than get back to the stable 2.6.32-4?

 

[tom]

see http://forum.proxmox.com/threads/70...el-panic-after-uping-to-1.9?p=40129#post40129

 

[rengiared]

ok thx, and its definitely possible to get back to 2.6.32-4?

 

[anullinger]

Hi, i think i got the same Problem here.

Updated Yesterday to 1.9 with Kernel 32-6. Proxmox Mail Gateway 2.6 won't startup anymore. 100% CPU Load with 2 vCPU.
If i change to 1 vCPU it start and it look like it's working, no further tests carried out because it's a production system.
Two Windows 2008R2 Server with 4 and 8 vCPU work like a charme.

pve-manager: 1.9-24 (pve-manager/1.9/6542)
running kernel: 2.6.32-6-pve
pve-kernel-2.6.32-4-pve: 2.6.32-33
pve-kernel-2.6.32-6-pve: 2.6.32-43
qemu-server: 1.1-32
pve-firmware: 1.0-13
libpve-storage-perl: 1.0-19
vncterm: 0.9-2
vzctl: 3.0.28-1pve5
vzdump: 1.2-15
vzprocps: 2.0.11-2
vzquota: 3.0.11-1

regards Tom

 

[rengiared]

"nice" to hear (i hope you unterstand how i mean that), how have you distributed the vCPUs? All 4 as Socket, or 1 Socket und 4 Cores, or maybe 2 Sockets und 2 Cores and so on.

in the mailgateway i have 2 Sockets with each 1 core.
in 2008r2 i had the problem with 1 socket and 2 cores, 2 sockets and 2 cores

 

[anullinger]

"nice" to hear (i hope you unterstand how i mean that),

for sure

that's interesting

2 CPU Sockets, 1 Core: CPU 100 % and startup hold at "free unused memory" after every reset.
1 CPU Socket, 2 Cores: it get's further, mostly at "setting system clock" it reaches 100% CPU Load and hang, sometimes it can boot.
1 CPU Socket, 1 Core: no problems

dmesg at host says no errors

Strange, never seen this befor.

 

[tom]

your package list is incomplete, pls follow the upgrade howto:
http://pve.proxmox.com/wiki/Downloads#Update_a_running_Proxmox_Virtual_Environment_to_1.9

 

[tom]

post your 'pveversion -v'

 

[anullinger]

Ok there was a failure in manual paket update (there was no internet connection at time of upgrade)
Now pveversion is ok.

prox02:~# pveversion -v
pve-manager: 1.9-24 (pve-manager/1.9/6542)
running kernel: 2.6.32-6-pve
proxmox-ve-2.6.32: 1.9-47
pve-kernel-2.6.32-4-pve: 2.6.32-33
pve-kernel-2.6.32-6-pve: 2.6.32-47
qemu-server: 1.1-32
pve-firmware: 1.0-14
libpve-storage-perl: 1.0-19
vncterm: 0.9-2
vzctl: 3.0.29-2pve1
vzdump: 1.2-16
vzprocps: 2.0.11-2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.15.0-1
ksm-control-daemon: 1.0-6

Newest 1.9 Version from 27.Sep. but the Problem with Mail Gateway an 2 CPU Sockets still persists.

mfg
tom, too