VirtualBox

Ticket #6559 (closed defect: fixed)

Opened 4 years ago

Last modified 3 years ago

Guest OS Loses Network Connection

Reported by: jwilliams108 Owned by:
Priority: critical Component: network/hostif
Version: VirtualBox 3.2.12 Keywords:
Cc: Guest type: Linux
Host type: Mac OS X

Description

Running VirtualBox 3.1.6 on a Mac OS X Server 10.6.3 Host on an XServe, with CentOS 5.4 Guest. The Guest OS loses its network connection (both host-only and nat) within 5-10 minutes of startup. The network appears up, but I cannot access the Guest from the Host, nor can I get out from the Guest to the internet, lan or host machine. Restarting networking within the Guest does not fix the problem, nor does restarting the Guest OS. Saving the VM state and starting it again seems to be the only means of re-establishing network connectivity. I have also noticed, however, that continuous network access (i.e., pinging the guest) seems to keep it accessible. Strangely, this same VM setup works fine on 10.6.3 non-server on a Mac Pro.

Attachments

VBox.log Download (40.7 KB) - added by jwilliams108 4 years ago.
VBox.log-r62212.log Download (177.3 KB) - added by jwilliams108 4 years ago.
VBox.log after replacing VBoxDD.dylib
screenshot.png Download (264.2 KB) - added by jwilliams108 4 years ago.
Screenshot with 'Unable to load R3 module' error
nat.log Download (70.6 KB) - added by jwilliams108 4 years ago.
nat.2.log Download (66.9 KB) - added by jwilliams108 4 years ago.
nat.log with only one NAT adapter

Change History

comment:1 Changed 4 years ago by jwilliams108

Here is a forum thread on the subject:

 http://forums.virtualbox.org/viewtopic.php?t=29884

comment:2 Changed 4 years ago by Hachiman

The logs files with host-only and nat networking will be helpful to clarify what have happened.

comment:3 Changed 4 years ago by frank

So to summarize, this problem occurs on Mac OS X Server 10.6.3. Maybe indeed host-specific.

Changed 4 years ago by jwilliams108

comment:4 Changed 4 years ago by jwilliams108

I've attached a VBox.log file. The network connection dropped approximately 7 minutes after starting the virtual machine, but nothing appears to have been logged.

comment:5 follow-up: ↓ 6 Changed 4 years ago by jwilliams108

I compared logs from the same virtual machine installed on 10.6.3 non-server, and 10.6.3 server. The only difference is that the server log contains a number of lines similar to:

00:11:20.328 TM: Giving up catch-up attempt at a 60 000 056 151 ns lag; new total: 36 960 081 026 779 ns

But nothing like this appears in the non-server log. It seems the VM on the server os is gradually losing time. I'm not sure if this is at all related to the network issue.

comment:6 in reply to: ↑ 5 ; follow-up: ↓ 7 Changed 4 years ago by Hachiman

Replying to jwilliams108: Have you got a chance to check if behavior is the same for Leopard server? Or it specific for SL Server?

comment:7 in reply to: ↑ 6 ; follow-up: ↓ 8 Changed 4 years ago by jwilliams108

Replying to Hachiman: Sorry, I only have access to SL Server.

comment:8 in reply to: ↑ 7 ; follow-up: ↓ 9 Changed 4 years ago by Hachiman

Replying to jwilliams108:

Replying to Hachiman: Sorry, I only have access to SL Server.

Could you please check if 3.2 b2 works as it should on SL server? Probably it's indeed timer's issue.

comment:9 in reply to: ↑ 8 ; follow-up: ↓ 10 Changed 4 years ago by jwilliams108

Replying to Hachiman:

Could you please check if 3.2 b2 works as it should on SL server? Probably it's indeed timer's issue.

The network problem still exists in 3.2.0_BETA2, however, the 'Giving up catch-up attempt' errors seem to have been resolved.

comment:10 in reply to: ↑ 9 Changed 4 years ago by Hachiman

Replying to jwilliams108:

Replying to Hachiman:

Could you please check if 3.2 b2 works as it should on SL server? Probably it's indeed timer's issue.

The network problem still exists in 3.2.0_BETA2, however, the 'Giving up catch-up attempt' errors seem to have been resolved.

thanks for update.

comment:11 follow-up: ↓ 12 Changed 4 years ago by jwilliams108

Upgraded to 3.2.0, and the problem still exists.

comment:12 in reply to: ↑ 11 Changed 4 years ago by Hachiman

Replying to jwilliams108:

Upgraded to 3.2.0, and the problem still exists.

thanks for update. we're still expecting SL server distro here.

comment:13 follow-up: ↓ 15 Changed 4 years ago by Hachiman

  • Component changed from network to network/NAT

Could you please download  VBoxDD.dylib.6559 and replace corresponding library on your VBox installation?

# export VBOX_LOG=drv_nat.e.l2
# export VBOX_LOG_DEST=file=nat.log
# VirtualBox -startvm <your-vm-name>

Could you please attach result log to the ticket?

comment:14 Changed 4 years ago by Hachiman

  • Version changed from VirtualBox 3.1.6 to VirtualBox 3.2.0

comment:15 in reply to: ↑ 13 ; follow-up: ↓ 16 Changed 4 years ago by jwilliams108

Replying to Hachiman:

Could you please download  VBoxDD.dylib.6559 and replace corresponding library on your VBox installation?

I tried this, but receive an error on VM startup:

Unable to load R3 module
/Applications/VirtualBox.app/Contents/MacOS/VBoxDD.dylib (VBoxDD) (VERR_FILE_NOT_FOUND)

However, the file does exist. It works fine again if I replace it with the original VBoxDD.dylib.

comment:16 in reply to: ↑ 15 ; follow-up: ↓ 17 Changed 4 years ago by Hachiman

Replying to jwilliams108:

Replying to Hachiman:

Could you please download  VBoxDD.dylib.6559 and replace corresponding library on your VBox installation?

However, the file does exist. It works fine again if I replace it with the original VBoxDD.dylib.

Have you renamed downloaded VBoxDD.dylib.6559 -> VBoxDD.dylib before trying, right?

comment:17 in reply to: ↑ 16 Changed 4 years ago by jwilliams108

Replying to Hachiman:

Have you renamed downloaded VBoxDD.dylib.6559 -> VBoxDD.dylib before trying, right?

Yes. Also tried it on my Mac Pro with the same result.

comment:18 follow-up: ↓ 19 Changed 4 years ago by Hachiman

Could you please try again?  VBoxDD.dylib?

With the same environment variables?

comment:19 in reply to: ↑ 18 ; follow-up: ↓ 20 Changed 4 years ago by jwilliams108

Replying to Hachiman:

Could you please try again?  VBoxDD.dylib?

With the same environment variables?

Same result... 'Unable to load R3 module'. I upgraded to 3.2.2, and then tried this library, and although it runs, it gives a Guru Meditation error when restoring the virtual machine.

comment:20 in reply to: ↑ 19 ; follow-up: ↓ 21 Changed 4 years ago by Hachiman

Replying to jwilliams108:

Replying to Hachiman:

Could you please try again?  VBoxDD.dylib?

With the same environment variables?

Same result... 'Unable to load R3 module'. I upgraded to 3.2.2, and then tried this library, and although it runs, it gives a Guru Meditation error when restoring the virtual machine.

That possible... because in debug library might save/restore more information. So what I can recommend do not restore the state just let vm to boot in normal way.

Changed 4 years ago by jwilliams108

VBox.log after replacing VBoxDD.dylib

comment:21 in reply to: ↑ 20 ; follow-up: ↓ 22 Changed 4 years ago by jwilliams108

Replying to Hachiman:

So what I can recommend do not restore the state just let vm to boot in normal way.

Tried this, but still a Guru Meditation error. I've attached the log file for reference.

comment:22 in reply to: ↑ 21 ; follow-up: ↓ 23 Changed 4 years ago by Hachiman

Replying to jwilliams108:

Replying to Hachiman:

So what I can recommend do not restore the state just let vm to boot in normal way.

Tried this, but still a Guru Meditation error. I've attached the log file for reference.

Hm ..., ok I'll put release library with enabled logs.

comment:23 in reply to: ↑ 22 ; follow-up: ↓ 24 Changed 4 years ago by Hachiman

Replying to Hachiman:

Replying to jwilliams108:

Replying to Hachiman:

So what I can recommend do not restore the state just let vm to boot in normal way.

Tried this, but still a Guru Meditation error. I've attached the log file for reference.

Hm ..., ok I'll put release library with enabled logs.

I've uploaded  release bits, so it shouldn't be any guru for you. Could you please try with these bits? It gives less logs but at least it should produce events trace to the release log file. So you can run your vm in usual way just attach new log from vm.

comment:24 in reply to: ↑ 23 ; follow-up: ↓ 25 Changed 4 years ago by jwilliams108

Replying to Hachiman:

I've uploaded  release bits, so it shouldn't be any guru for you. Could you please try with these bits?

OK, I seem to be doing something wrong. I am getting the R3 module error again with this dylib. I've attached a screen shot displaying the error, and showing the file listing.

Changed 4 years ago by jwilliams108

Screenshot with 'Unable to load R3 module' error

comment:25 in reply to: ↑ 24 Changed 4 years ago by Hachiman

Replying to jwilliams108:

Replying to Hachiman:

I've uploaded  release bits, so it shouldn't be any guru for you. Could you please try with these bits?

OK, I seem to be doing something wrong. I am getting the R3 module error again with this dylib. I've attached a screen shot displaying the error, and showing the file listing.

Hm ... Let's do the following: please restore your original dylib.

# export VBOX_RELEASE_LOG=drv_nat.e.l2
# export VBOX_RELEASE_LOG_DEST=file=nat.log
# VirtualBox -startvm <your-vm>

I'd hoped that will give enough information to clarify what happens.

Changed 4 years ago by jwilliams108

comment:26 Changed 4 years ago by jwilliams108

I've attached the nat.log. The machine ran for about 20 minutes, and the networking stopped responding after about 10 minutes or so. Please let me know if you need more.

comment:27 follow-up: ↓ 28 Changed 4 years ago by Hachiman

Could you please configure your VM to have only one adapter enabled with NAT attachment and attach result log to ticket?

Changed 4 years ago by jwilliams108

nat.log with only one NAT adapter

comment:28 in reply to: ↑ 27 ; follow-up: ↓ 29 Changed 4 years ago by jwilliams108

Replying to Hachiman:

Could you please configure your VM to have only one adapter enabled with NAT attachment and attach result log to ticket?

I've attached the new log... note that I ran the machine for about 40 minutes, and could still reach the internet from the virtual machine. Maybe the problem is host-only networking specific?

comment:29 in reply to: ↑ 28 ; follow-up: ↓ 30 Changed 4 years ago by Hachiman

Replying to jwilliams108:

Replying to Hachiman:

Could you please configure your VM to have only one adapter enabled with NAT attachment and attach result log to ticket?

I've attached the new log... note that I ran the machine for about 40 minutes, and could still reach the internet from the virtual machine. Maybe the problem is host-only networking specific?

'host-only networking' means ability to access to host from the guest only.

comment:30 in reply to: ↑ 29 Changed 4 years ago by jwilliams108

Replying to Hachiman:

'host-only networking' means ability to access to host from the guest only.

Yes, my setup is 2 interfaces, 1 - host-only, for access to that vm directly from the OS X Server host (which also acts as a public web proxy for the private vms) or via VPN; and, 2 - nat for external internet access for each vm.

So, the main problem is that this host <-> guest communication fails after 10 minutes or so unless I have a continuous ping occuring. Previously, the external internet access also failed, but this seems to be working properly now.

comment:31 follow-up: ↓ 33 Changed 4 years ago by frank

Any chance to retest with VBox 3.2.6 Beta 2?

comment:32 Changed 4 years ago by Hachiman

  • Component changed from network/NAT to network/hostif

comment:33 in reply to: ↑ 31 Changed 4 years ago by jwilliams108

Issue still present in 3.2.6 Beta 2.

comment:34 follow-up: ↓ 36 Changed 4 years ago by jwilliams108

Issue still present in 3.2.6 release.

comment:35 Changed 4 years ago by buffyg

Have you compared this with 3783 and tried setting the MTU to 1468? I see that you're also reporting problems with NAT, which may mean this is something else again. Not sure about CentOS having MTU problems, either, as I've noticed that not all guests seem to be inhibited by the MTU issue (e.g. OpenBSD and Ubuntu seem to handle it alright, whereas OpenSolaris and Windows not so much).

comment:36 in reply to: ↑ 34 Changed 4 years ago by Hachiman

Replying to jwilliams108:

Issue still present in 3.2.6 release.

Does it happen in combination NAT + Host-Only or can be achieved with Host-Only? Could you please dump guest route information on your SL and SL server hosts, might be caused by some guest misconfiguration?

comment:37 Changed 4 years ago by jwilliams108

Issue still present in 3.2.8. The problem results with NAT + Host-only and Bridged + Host-only. I'll check with just Host-only when I have a chance. The routes are fine - as I mentioned, everything works well as long as a ping keeps the network alive. Also, I have noted that when I run two guest machines, I only need to ping one to keep the host-only networks alive on both.

comment:38 Changed 4 years ago by slobo

I am experiencing this issue on Windows XP host / Ubuntu 10.04 guest. I have not verified that continuous ping keeps the connection open.

comment:39 follow-up: ↓ 40 Changed 4 years ago by frank

Guys, could you be more specific? Does this problem occur as well if only a host-only network is assigned to the VM or is it necessary to assign a NAT network to the VM as well to reproduce this bug?

comment:40 in reply to: ↑ 39 Changed 4 years ago by jwilliams108

Replying to frank:

Guys, could you be more specific? Does this problem occur as well if only a host-only network is assigned to the VM or is it necessary to assign a NAT network to the VM as well to reproduce this bug?

I've been able to test host-only, and the issue does not appear when using host-only. It seems to required an additional NAT or Bridged adapter.

comment:41 follow-up: ↓ 42 Changed 3 years ago by roger_p_melet

Hi Guys,

I've been experiencing this issue also on host: MacOSX server and guest: WindowsXP, Debian and Ubuntu. In "Network" I've selected adapter type: Intel PRO/1000MT Desktop (82540EM). Lost network connection within 10 minutes.

Now I've change the adapter type to: Intel PRO/1000 T Server (82543GC) and it takes about 1 hour before the connection is lost. It looks like its lost because ping times out or ping time is more than 20000.000 ms.

I'm going to try the next adatpter

comment:42 in reply to: ↑ 41 Changed 3 years ago by swingkid

Replying to roger_p_melet:

Hi Guys,

I've been experiencing this issue also on host: MacOSX server and guest: WindowsXP, Debian and Ubuntu. In "Network" I've selected adapter type: Intel PRO/1000MT Desktop (82540EM). Lost network connection within 10 minutes.

Now I've change the adapter type to: Intel PRO/1000 T Server (82543GC) and it takes about 1 hour before the connection is lost. It looks like its lost because ping times out or ping time is more than 20000.000 ms.

I'm going to try the next adatpter

I just wanted to let people know that with my setup. Host: Windows 7 Ultimate (64 bit) running VBOX 3.2.6 r63112 and Guest (Windows XP sp3), I have been losing guest network connectivity. I also have two nic cards configured one for Host ONly the other for NAT.

The connectivity sometimes drops when I try and surf a website. Other itmes when I try and connect up a running jboss to connect to an external database. I was running continous pings to an external host so I know the continous pings didn't seem to help keep the ntwork connectivity alive.(at least in my case).

Just wanted to add a datapoint.

comment:43 Changed 3 years ago by languste

hi,

i'm facing the same problem running VirtualBox 3.2.10 on Mac OS X 10.6.5 with different guests (bridged-networking).
here my observations/tests:

  • continuous ping solves it
  • changing from DHCP to static IP (inside guest OS) solves it

my guess is a bug involving broadcast packet handling.

and by the way i remeber vmware having the same issue years ago.

cheers, martin

comment:44 follow-up: ↓ 45 Changed 3 years ago by kelm

Same thing here, running Windows Server 2008 and CentOS 5.5 VMs on a Mac OS X Server 10.6.5 host. Both VMs are connected in bridged mode.

I cannot confirm Martin's observations. Both machines here have a static IP and pings did not resolve the issue entirely, especially on the Windows host. Pinged the VMs every two minutes continuously plus pinged an external machine from (inside) the VMs every minute. The Windows Server host still lost networking repeatedly.

Interestingly, I got different results with different virtual NIC models. The Intel Pro 1000 MT desktop lost connectivity much earlier than the 1000 MT Server. It seems that we can use the ping workaround with the latter (for now) to maintain connectivity.

What info could we provide to help get this nasty bug closed?

Peter

comment:45 in reply to: ↑ 44 Changed 3 years ago by kelm

Oops - forgot to mention this: VirtualBox version is 3.2.12.

comment:46 Changed 3 years ago by frank

  • Version changed from VirtualBox 3.2.0 to VirtualBox 3.2.12

comment:47 Changed 3 years ago by wolfgangihloff

I experience the same on CentOS host with Guest being Windows XP 32bit Prof, however only one of the three VM Guests has this problem.

comment:48 Changed 3 years ago by roger_p_melet

I experience no more problems with version 4.0.0 r 69151.

comment:49 Changed 3 years ago by frank

  • Status changed from new to closed
  • Resolution set to fixed

Please reopen if still relevant with VBox 4.0.6.

Note: See TracTickets for help on using tickets.

www.oracle.com
ContactPrivacy policyTerms of Use