Ubuntu Intrepid / RHEL 5 Guest / VBox 2.0.2 guest crashes

Discussions related to using VirtualBox on Linux hosts.
Post Reply
vondralbra
Posts: 15
Joined: 22. Oct 2008, 10:31
Primary OS: Debian Lenny
VBox Version: PUEL
Guest OSses: Linux

Ubuntu Intrepid / RHEL 5 Guest / VBox 2.0.2 guest crashes

Post by vondralbra »

Dear audience,


there seems to be a bug in the virtual memory management (or a bug in the settings of my virtualbox environment).

Each time when the RHEL-Guest is a bit burdened the virtual machine enters the Guru-Meditation.
(frankly: Each time during the installation of an Oracle Database on the RHEL-Guest)

The log shows reproduceably the following error:

Code: Select all

00:02:36.064 Guest Log: vboxadd_hgcm_connect: Connected, client ID 2
             
             !!Assertion Failed!!
             Expression: RT_SUCCESS_NP(rc)
             Location  : /home/vbox/vbox-2.0.2/src/VBox/VMM/VMMAll/PGMAllPool.cpp(1517) int pgmPoolMonitorInsert(PGMPOOL*, PGMPOOLPAGE*)
             -1701
00:25:23.971 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
00:25:23.971 !!
00:25:23.971 !!                 Guru Meditation 1103 (VINF_EM_DBG_HYPER_ASSERTION)
00:25:23.971 !!
00:25:24.120 
00:25:24.120 !!Assertion Failed!!
00:25:24.120 Expression: RT_SUCCESS_NP(rc)
I already searched the forum and the bugtracker but I only found those three links.

http://forums.virtualbox.org/viewtopic. ... itorinsert
http://forums.virtualbox.org/viewtopic. ... itorinsert
and finally
http://www.virtualbox.org/ticket/1629

Neither one is containing a solution.
Are there any ideas? I'm pretty much out of them.


Best regards

vda
vondralbra
Posts: 15
Joined: 22. Oct 2008, 10:31
Primary OS: Debian Lenny
VBox Version: PUEL
Guest OSses: Linux

digging the source of the crash....

Post by vondralbra »

I digged a bit in the sources and revealed the meaning of the "1701" error code.

/** Hypervisor memory allocation failed. */
#define VERR_MM_HYPER_NO_MEMORY (-1701)


it all breaks down to a call to MMHyperAlloc() which returns this errorcode on two occasions where one is masked by a #ifdef rule which means that there should be a message in the log stating:

Code: Select all

    Log2(("MMHyperAlloc: cb=%#x uAlignment=%#x returns VERR_MM_HYPER_NO_MEMORY\n", cb, uAlignment));
    AssertMsgFailed(("Failed to allocate %d bytes!\n", cb));

    return VERR_MM_HYPER_NO_MEMORY;
(which isn't - or at least is suppressed by some config option)

But the main issue here is: This would mean that there is not enough memory to run a virtual machine using 1.5G RAM? Am I right?

Best regards

vda
sandervl
Volunteer
Posts: 1064
Joined: 10. May 2007, 10:27
Primary OS: MS Windows Vista
VBox Version: PUEL
Guest OSses: Windows, Linux, Solaris

Post by sandervl »

How much of the 1.5 GB host RAM did you assign to the VM?
vondralbra
Posts: 15
Joined: 22. Oct 2008, 10:31
Primary OS: Debian Lenny
VBox Version: PUEL
Guest OSses: Linux

Post by vondralbra »

sandervl wrote:How much of the 1.5 GB host RAM did you assign to the VM?
The host machines RAM is (no VM started)

Code: Select all

             total       used       free     shared    buffers     cached
Mem:       4046292     794596    3251696          0      21816     347596
-/+ buffers/cache:     425184    3621108
Swap:      6016300          0    6016300
and like this

Code: Select all

             total       used       free     shared    buffers     cached
Mem:       4046292    2227448    1818844          0      28512     969584
-/+ buffers/cache:    1229352    2816940
Swap:      6016300          0    6016300
when the 1.5G sized RHEL5 Guest is freshly booted and idling along.

During the installation the RAM is:

Code: Select all

             total       used       free     shared    buffers     cached
Mem:       4046292    4017020      29272          0      29628    1764896
-/+ buffers/cache:    2222496    1823796
Swap:      6016300       2440    6013860
At the time of crash the host's RAM is like this:

Code: Select all

             total       used       free     shared    buffers     cached
Mem:       4046292    4020896      25396          0      23772    1765640
-/+ buffers/cache:    2231484    1814808
Swap:      6016300       5424    6010876
The crashed machine's process is

Code: Select all

ps xufa | grep orael32
ta        7075 86.4 40.9 1949468 1655804 ?     Sl   15:11  14:54  \_ /usr/lib/virtualbox/VirtualBox
The guest crashes all the same when set to 900M, 1024M and 1.5G.
If logs or any other debugging measures are needed I'll gladly provide them.

Best regards

vda
sandervl
Volunteer
Posts: 1064
Joined: 10. May 2007, 10:27
Primary OS: MS Windows Vista
VBox Version: PUEL
Guest OSses: Windows, Linux, Solaris

Post by sandervl »

That looks like a memory leak of some kind. Could you create a defect on our bug tracker? Also include the VBox.log of the failed session.

Thanks.
vondralbra
Posts: 15
Joined: 22. Oct 2008, 10:31
Primary OS: Debian Lenny
VBox Version: PUEL
Guest OSses: Linux

Post by vondralbra »

sandervl wrote:That looks like a memory leak of some kind. Could you create a defect on our bug tracker? Also include the VBox.log of the failed session.

Thanks.
Thank you very much for the permission to enter a ticket into the bugtracker.
If there are any remote hands needed to debug the issue on site feel free to ask.
I'll gladly help.

Best regards.

vda
vondralbra
Posts: 15
Joined: 22. Oct 2008, 10:31
Primary OS: Debian Lenny
VBox Version: PUEL
Guest OSses: Linux

Reducing Machine size solved the crash

Post by vondralbra »

I found a workaround to solve this issue for me: I reduced the Guest-Machines size to 768M. Oracle complained about too little RAM (922M needed but 768M supplied) but the installer went through without any errors and the instance came up.

But this error should not happen anyway - and if it happened it would be fine to have a speaking errormessage in the logfile at least.

Best regards.

vda
fixedwheel
Volunteer
Posts: 1699
Joined: 13. Sep 2008, 02:18

Re: Reducing Machine size solved the crash

Post by fixedwheel »

vondralbra wrote:workaround to solve this issue for me: I reduced the Guest-Machines size to 768M.
the 32bit linux kernel lowmem/highmem boundary is at 896M, AFAIK

i bet you can safely add another 128M to the 768M ;)
vondralbra
Posts: 15
Joined: 22. Oct 2008, 10:31
Primary OS: Debian Lenny
VBox Version: PUEL
Guest OSses: Linux

Re: Reducing Machine size solved the crash

Post by vondralbra »

fixedwheel wrote:
vondralbra wrote:workaround to solve this issue for me: I reduced the Guest-Machines size to 768M.
the 32bit linux kernel lowmem/highmem boundary is at 896M, AFAIK

i bet you can safely add another 128M to the 768M ;)
Not quite. I tried to lift the guest's RAM to 896M and in fact the first two hours went fine during the patch of 10.2.0.1 to 10.2.0.4 but then this was the result:

Code: Select all

02:52:53.282 PATM: Disable block at c06fedd7 - invalid write c06fede0-c06fede4 
             
             !!Assertion Failed!!
             Expression: RT_SUCCESS_NP(rc)
             Location  : /home/vbox/vbox-1.6.6/src/VBox/VMM/VMMAll/PGMAllPool.cpp(1310) int pgmPoolMonitorInsert(PGMPOOL*, PGMPOOLPAGE*)
             -1701
03:26:26.740 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
03:26:26.740 !!
03:26:26.740 !!                 Guru Meditation 1103 (VINF_EM_DBG_HYPER_ASSERTION)
03:26:26.740 !!
This "PATM: Disable block at ..." makes me wonder whether the hardware is as healthy as it should be.


Best regards.

vda
sandervl
Volunteer
Posts: 1064
Joined: 10. May 2007, 10:27
Primary OS: MS Windows Vista
VBox Version: PUEL
Guest OSses: Windows, Linux, Solaris

Re: Reducing Machine size solved the crash

Post by sandervl »

vondralbra wrote:

Code: Select all

02:52:53.282 PATM: Disable block at c06fedd7 - invalid write c06fede0-c06fede4 
This "PATM: Disable block at ..." makes me wonder whether the hardware is as healthy as it should be.
vda
You can ignore that warning. It's harmless.
sej7278
Volunteer
Posts: 1003
Joined: 5. Sep 2008, 14:40
Primary OS: Debian other
VBox Version: PUEL
Guest OSses: Solaris, Linux, Windows, OS/2, MacOSX, FreeBSD
Contact:

Post by sej7278 »

dunno if this is much help but this week i have successfully installed oracle 10gr2, 9ir2 and 11gr1 under centos 5.2 and 4.7 (i.e. rhel 5u2/4u7) on virtualbox 2.0.2

32-bit guests, fedora9 64-bit host (core2quad). 1gb allocated to the vm's, 8gb host ram. bridged networking. all latest kernels/updates etc.

i saw no issues at all from virtualbox, although i did do all the strange oracle-recommended memory settings and kernel parameters in the guests before installing.

the only time i've ever seen a problem when the guest gets a bit loaded is when i've run vbox over an ssh-forwarded x11 session, then the guest and vbox process just dies with some SDL error in the logs, happened with winxp and centos guests.
vondralbra
Posts: 15
Joined: 22. Oct 2008, 10:31
Primary OS: Debian Lenny
VBox Version: PUEL
Guest OSses: Linux

Post by vondralbra »

sej7278 wrote:dunno if this is much help but this week i have successfully installed oracle 10gr2, 9ir2 and 11gr1 under centos 5.2 and 4.7 (i.e. rhel 5u2/4u7) on virtualbox 2.0.2
32-bit guests, fedora9 64-bit host (core2quad). 1gb allocated to the vm's, 8gb host ram. bridged networking. all latest kernels/updates etc.
i saw no issues at all from virtualbox, although i did do all the strange oracle-recommended memory settings and kernel parameters in the guests before installing.
Undoubtly there is no way to install Oracle without those memory settings. I do not think that they are responsible for this crash neither. Yesterday evening I reduced the guest's RAM to 768M again, reinstalled 10.2.0.1 (disregarding the low RAM warning) and patched up to 10.2.0.4 without any hassles where a Guest having 896M of RAM crashed. How did you set up the Disks? I followed the OFA by having four virtual disks connected via SATA to the Guest - /, /u01/apps, /u02/oradata and /u03/oradata.

The host in question here is a Notebook having 4G RAM and a Dual-CPU and Oracle in a VBox is just for educational purposes.
the only time i've ever seen a problem when the guest gets a bit loaded is when i've run vbox over an ssh-forwarded x11 session, then the guest and vbox process just dies with some SDL error in the logs, happened with winxp and centos guests.
This would have been one of my next steps to get a desktop of the virtual machine onto the screen on my desk (Notebook keyboards are a bit inconvenient)


Best regards


vda
sej7278
Volunteer
Posts: 1003
Joined: 5. Sep 2008, 14:40
Primary OS: Debian other
VBox Version: PUEL
Guest OSses: Solaris, Linux, Windows, OS/2, MacOSX, FreeBSD
Contact:

Post by sej7278 »

i used these instructions for 11g and 10g:

http://www.idevelopment.info/data/Oracl ... X_20.shtml

http://www.idevelopment.info/data/Oracl ... X_15.shtml

and kind of googled a lot to get 9i working, its reliant on some pretty old software these days - and i had to use ide due to 2.6.9 kernel on centos 4.7

my 10g/11g setup differs in that i had one sata drive with /u01, /u02 etc as partitions, not optimal performance, but it is only for development. maybe virtualbox isn't keen on using 4+ drives.....?

as far as x11-over-ssh goes, its not a constant issue, but i have seen it a lot - its about my only problem with virtualbox at the moment, there's something up with the SDL console, i guess the alternative is to use VBoxHeadless to start vm's and RDP to access the GUI, but thats slower - or ssh into the guest and export x11 from that, then you don't have control of the vbox application though.
Post Reply