VBox VM's choking/freezing on ata1.00 errors Centos 5.5 on Centos 5.5

[Billy]

Hey There, Using Dedicated server with

• x2 Quad Core Xeon 5600 w/ 12 Gb RAM, and 1Tb Sata HD. Running Centos 5.5

VBox VM: VBox machine running with

• Centos 5.5, 1 Core, 1024 MB RAM, SATA, 20 Gb Growing or Fixed.

Symptoms: VM running tickety boo, vrdp view pukes out error (See attached) Able to hit enter to get to login, and the server freezes. If logged in at the time, then able to run one command then freezes. If left alone, like in top, then it runs for a while and chokes.

Tried 3.2.4 but got all kinds of errors relating to Network. Didn't want to cross that bridge so I went back to 3.2.2

Errors seem related to Hard Drive inaccessibility. But not able to pin it down. I have 10 sessions of the same server running (for testing) and they stay alive. Happens when idle, or being used. Not Predicatable.

Checked vbox.log and no info Checked Dmesg and no info Checked dmesg on VM and no info

VBoxManage controlvm poweroff

results are: 0%...10%...20%... and freezes. Need to Ctrl C and kill pid

Restart and the VM machine will be fine for a while. Predictable that it will happen again.

[Billy]

Screencap of VRDP Error Puke of server A

[Billy]

Screencap of VRDP Error Puke of server B

[aeichner]

Please attach a VBox.log.

[Billy]

Log File of VM that had just crashed

[Billy]

More VRDP SCreen Dump 1 same crash as Log File

[Billy]

More VRDP Screen Dump 2 same crash as Log File

[Billy]

I sent one of these crashed machines a reset command this is what I got back

# VBoxManage controlvm Planner_Courtenay reset Oracle VM VirtualBox Command Line Management Interface Version 3.2.2 (C) 2005-2010 Oracle Corporation All rights reserved.

ERROR: Could not reset the machine (VERR_VM_INVALID_VM_STATE) Details: code VBOX_E_VM_ERROR (0x80bb0003), component Console, interface IConsole, callee nsISupports Context: "Reset()" at line 112 of file VBoxManageControlVM.cpp

[Billy]

Also I update to 3.2.4 on my second server, and it is getting these frozen states as well

[Billy]

Log after attempting to reset a frozen state machine

[Billy]

May have found answer on IRC Here is Dialog:

<BillyStar>    My problem is the VMs are freezing after a while with no good explanation
<BillyStar>    I've scoured the logs and everything seems fine
<BillyStar>    The VM is a Default Centos 5.5 runlevel 3 on a server with Centos 5.5 runlevel 3
<BillyStar>    The only clue is a screen dump from vrdp screens
<BillyStar>    http://www.virtualbox.org/attachment/ticket/6955/VMError1.jpg
<klaus-vb>    which vbox version, which kind of storage controller?
<klaus-vb>    which linux version?
<BillyStar>    VBox 3.2.2 and 3.2.4
<BillyStar>    Centos 5.5
<Voldenet>    BillyStar, is this netinst?
<BillyStar>    Used Yum install VirtualBox-3.2
<BillyStar>    Not sure which controller. It's in proc somewhere...?
<BillyStar>    sata. I know that at least
<sol>    BillyStar; did you upgrade from 3.1.x or just fresh install 3.2.x?
<BillyStar>    Fresh 3.2
<BillyStar>    I can restart the VM's and they run okay for a bit. Not heavy usage, but they will eventually freeze in the same way
<BillyStar>    The VBox Server is stable. No Errors
<BillyStar>    I tried static and fixed hd images
<klaus-vb>    BillyStar: what filesystem on the host?
<sol>    BillyStar; if you tail -f your VM log, does anything occur when the freeze happens?
<BillyStar>    Filesystem ext3
<BillyStar>    Nothing in the log. If I try controlvm reset, then I get an invalid state err in the log
<klaus-vb>    BillyStar: try enabling the host cache with VBoxManage storagectl... linux async I/O turned out to be not very solid, so maybe it's worth trying without.
<BillyStar>    Will do. I was reading about that but wasn't familiar territory.
<BillyStar>    Isn't the cache enabled by default?
<klaus-vb>    with 3.2.x it is enabled by default only for ide and floppy. sata/scsi have it disabled, as it allows using multiple outstanding I/O operations, which is more of a performance boost than the host i/o cache shuld be...
<klaus-vb>    but many filesystems have bugs. ext4 can't handle async I/O for files > 4G, btrfs can't handle it at all and so on. all beta crap.
<tru_tru>    BillyStar: same setup, I also have random 'hang' even during the guest boot (guest has sata controler, a single expandable hd)
<BillyStar>    I generally can get up and running. Stable for about 5 minutes or so then it's a crap shoot
<tru_tru>    it's not that bad for me.. up 18:02
<sol>    hehe if I didn't know this conversation was about vbox it'd be funnier ;)
<BillyStar>    There is a joke in there somewhere about my wifes' VirtualBox
<BillyStar>    So I've enabled cache for my 10 VMs on one server and have restarted. Stay tuned
<tru_tru>    VBoxManage storagectl vm_name --name 'SATA Controller' --hostiocache {on|off} does not return any error :)
<tru_tru>    how do I check if hostiocache is indeed on or off?
<klaus-vb>    don't know though if {on|off} is interpreted as on or off :D
    * tru_tru    not clear, sorry: I try --hostiocache on ; then --hostiocache off
<BillyStar>    if you look at the vmname.xml file in .VirtualBox/Machines/ under SataController you can see if it is on
<klaus-vb>    oh :) it's not expected to say anything unless you made a mistake. unix style commands.
<klaus-vb>    yep. seems no one thought of putting that setting yet in the output of VBoxManage showvminfo
<tru_tru>    I was searching the VBoxManage man pages for displaying the current status
<tru_tru>    BillyStar: thanks (grepping useHostIOCache in the xml file)

[Billy]

Okay to close. Thank you.

Summary: Centos 5.5 VM on Centos Server with VirtualBox 3.2.2 or 3.2.4. VM's would hang and require restart. See attached Log Files and screen shots.

running headlesss vm's Resolved by enabling HostIOCache with VBoxManage StorageCtl vmname --hostiocache on