UNSOLVED
A
AlexSynapse
16 Posts
0
1100
August 17th, 2015 07:00
Failed SVM NICs after reboot
I currently have an issue with a secondary SVM - the server had been shutdown whilst an ESXi host was placed in maintenance mode. When the host was placed back in production, and the SVM powered on, it no longer finds any of the eth interfaces within the SVM. Please see screenshots below:
http://postimg.org/image/h3qjwkoyb/
http://postimg.org/image/8ps3xcsmh
The nics are connected within vsphere - any ideas where to start with this// why this has happened.
I've also had constant issues with rebooting any SVM. Nine times out of ten I will need to reboot the VM several times before it boots correctly (black screen, so no ideas why). At other times /dev/sda2 will show errors and require me to manually run e2fsck before it comes back online. This can't be normal!
This is on the latest version of ESX/ vSphere (6.0) and with ScaleIO 1.32.
Alex,
Quick check- Are the MAC addresses on the NICs changing each boot? (If you edit the VM and expand the NIC's properties, you'll see the MAC address.)
If not, it's more likely to be in the SVM's udev persistent naming rule or network script in /etc/sysconfig.
What config files have you compared between this SVM and the functioning ones?
AlexSynapse
16 Posts
719
0
Posted August 18th, 2015 02:00
The MAC addresses are the same after reboot.
I’m beginning to wonder whether it’s a vmware tools issue, as I also have an error about mounting the HGFS share (see first screen shot). Unfortunately, as it’s an appliance, I can’t simply reinstall tools to see if sorts the issue.
In terms of what I’ve been comparing between SVMs, it’s the usual suspects:
/etc/sysconfig/network/* (ifcg-eht*, configs, etc)
I can’t find any differences at all.
AlexSynapse
16 Posts
719
0
Posted August 18th, 2015 02:00
I forgot to add that the persistent net rules in udev look correct too – the mac addresses match the virtual NICs for each interface.
From: Alex Hudson
Sent: 18 August 2015 10:18
To: 'jive-1456666747-bhcx-2-jbyk@emc-ecn.hosted.jivesoftware.com'
Subject: RE: - Failed SVM NICs after reboot
The MAC addresses are the same after reboot.
I’m beginning to wonder whether it’s a vmware tools issue, as I also have an error about mounting the HGFS share (see first screen shot). Unfortunately, as it’s an appliance, I can’t simply reinstall tools to see if sorts the issue.
In terms of what I’ve been comparing between SVMs, it’s the usual suspects:
/etc/sysconfig/network/* (ifcg-eht*, configs, etc)
I can’t find any differences at all.
AlexSynapse
16 Posts
719
0
Posted August 18th, 2015 06:00
I’ve managed to get the VM to maintain the interfaces on reboot, but not in a way I’m particularly happy with as it now differs from the other SVMs.
After a reboot, I noticed via lsmod that the vmxnet3 driver wasn’t loaded. To get around it I edited the modules loaded at boot in /etc/sysconfig/kernel:
MODULES_LOADED_ON_BOOT=”vmxnet3”
The VM is now functioning again and can happily be rebooted without losing the interfaces.
I’d still like to figure out what has gone wrong and get this SVM back in line with the others. Any suggestions appreciated.
AlexSynapse
16 Posts
719
0
Posted August 21st, 2015 03:00
Yes they are.
AlexSynapse
16 Posts
719
0
Posted August 17th, 2015 10:00
I've managed to temporarily get around the issue by using yast to redo the interfaces; however, it is non-persistent. The config is lost again the next time I reboot the appliance.
Comparing the various config files against a another SVM hasn't highlighted any differences either. Everything looks identical.