VMware Cloud Community
stilmon
Contributor
Contributor

E1000 Having socket issues?

I seem to be noticing alot of socket issues with VM's connecting to even eachother or diffrent hosts. My layout is, I have only 10 VM's. My company's custom software takes in 8 million transactions a day and then sends it to another VM that decodes the information and writes the information to our sql database. Now my problem is my software is having trouble keeping or even establishing socket connections to eachother even if they are on the same host. THis did not happen until we went virtual. From research I have gathered it seems like the NIC E1000 driver may be at fault. It seems very easy to change the driver simply by deleting the current nic/driver (e1000) and adding a new one.  Now my other question is

Whats the way to go? E1000, VMXNET2 or VMXNET3

Also im running esxi 4 and vcenter 4.1.   Main switch has been replaced just to void that out being a problem.

Thanks!

0 Kudos
12 Replies
Troy_Clavell
Immortal
Immortal

If your guests have virtual hardware v7, I would go with VMXNET3

Choosing a network adapter for your virtual machine (1001805)

With that said, it could be your vSwitches

http://www.yellow-bricks.com/2010/02/02/e1000-and-dropped-rx-packets/

0 Kudos
stilmon
Contributor
Contributor

My VM is V7 I'll give that a try and let you know what happens later today. I appreciate your fast response!

0 Kudos
chriswahl
Virtuoso
Virtuoso

I use the VMXNET3 on all VMs / templates, haven't had any issues with it. It also shows itself as a 10GB connection in the guest.

VCDX #104 (DCV, NV) ஃ WahlNetwork.com ஃ @ChrisWahl ஃ Author, Networking for VMware Administrators
0 Kudos
stilmon
Contributor
Contributor

Wow 10GB eh? nice... So I'm assuming there is nothing special about doing this then. simply removing my network card (virtual) and then readding another one with the new drivers?

0 Kudos
Troy_Clavell
Immortal
Immortal

yeah VMXNET3 is a 10GB adapter.

For you Windows Guests, you may want to go  through and remove any hidden devices. Your old NIC may show up in  hidden devices.  For that matter anything that is grayed out should be uninstalled

http://support.microsoft.com/?kbid=269155

0 Kudos
chriswahl
Virtuoso
Virtuoso

That's it. VM Tools will provide the driver. If you have another hidden NIC there you'll get a message about having 2 gateways, you can remove the NIC or just ignore it.

VCDX #104 (DCV, NV) ஃ WahlNetwork.com ஃ @ChrisWahl ஃ Author, Networking for VMware Administrators
0 Kudos
stilmon
Contributor
Contributor

So probably a good Idea to do this to the Nic assigned to the Scsi? 

0 Kudos
stilmon
Contributor
Contributor

image.jpgWell i have made changes to the Nic cards and still same thing. Not sure what to do at this point. I have put in a request with VMware and they where unable to help so far. The forum is my last resort Smiley Sad It seems the VM's (all) will randomly lock up. usually a millisecond or 2. Here are some of the logs. I think maybe the storage array may be faulty? Not sure. please note the storage device connectivity.

[2011-09-26 21:39:24.332 00256 error 'App'] [Vdb::Connection::Commit] Connection commit failed: -1. Start retrying 20 more times.

an error in system log

0 Kudos
stilmon
Contributor
Contributor

Anyone?

0 Kudos
rajvm256
Enthusiast
Enthusiast

It seems to be a fault with the storage device, might be a path. Few things what I would have done. Check the vmkernel you might find the same errors.  possibly write a script to monitor the log and update when such a error occurs. Try to check if you were doing anything specific during that time. Is all your transactions happening at that time since you have already contacted the vmware support they should be best guys to help it Smiley Happy

Thanks | http://virtualvm.info/
0 Kudos
stilmon
Contributor
Contributor

Pretty sure I found the issue to be from my san. Now i made a mistake of removing a static target added by dynamic discovery and now I cannot get my esxi server to connect to the storage array. From what I read I must restart the server host. I have rescanned the ip's after they where added and still same thing. anyway to do a putty command instead of restarting host?

0 Kudos
stilmon
Contributor
Contributor

Well, restarting an old thread,

Whenever my vmhosts hvae a path to a lun thats been lost it causes my custom applications on windows server 2008 R2 that rely heavily on tcp/ip sockets to fail.

ex, if my application goes to open a socket to another application (thats on same host or vm) it fails, the applications have to retry multiple times 10-15 before they will get through... the only fix to this is to reinstall windows which is such a burden...  esxi 4.1  anybody have any insite? would be awsome!!!! thanks!!!....

0 Kudos