Description of problem: I have in oVirt defined a network called (and used as) SAN where also I have enabled Jumbo Frames. My Host physical interface em2 is hooked up to this net hence got an MTU of 9000. em2 Link encap:Ethernet HWaddr D0:67:E5:F9:2E:1C inet6 addr: fe80::d267:e5ff:fef9:2e1c/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:9000 Metric:1 RX packets:503209263 errors:0 dropped:0 overruns:0 frame:0 TX packets:483040537 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:2395173035170 (2.1 TiB) TX bytes:408352341106 (380.3 GiB) Still I see VM TAP interfaces attached to this bridge getting a default MTU of 1500, and therefore the NIC within the VM is anything but stable. Here is such a TAP interface, created when the VM starts, # Host Virtual Nic vnet26 Link encap:Ethernet HWaddr FE:1A:4A:2F:D2:A8 inet6 addr: fe80::fc1a:4aff:fe2f:d2a8/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:8 errors:0 dropped:0 overruns:0 frame:0 TX packets:6 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:500 RX bytes:648 (648.0 b) TX bytes:468 (468.0 b) If I manually set the MTU on that interface everything work as expected. The engine is currently on Beta2 but have been upgraded several times. This behaviour was introduced somewhere along the way, since I have several machines running all (still) have Jumbo Frames on the the interface attached to this bridge. Newly created VMs get 1500 though. I also found out just now that if I reboot a VM that too gets the incorrect MTU on that NIC. It's as if that MTU override is lost in space. Version-Release number of selected component (if applicable): Engine: 3.5.0-0.0.master.20140130172954.git9257b30.el6 Node: oVirt Node Hypervisor release 3.0.3 (0.999.201401231512draft.el6) How reproducible: Not sure :-) Try to do a MTU override on a network and attach a VM to it.
Hmmm, I just noticed the version number of the Engine. I have not seen that before, does it mean I have managed to add some nightly ? I was quite sure I only had added 3.4 repos.
[root@dashboard yum.repos.d]# rpm -qa | grep -i ovirt ovirt-engine-lib-3.5.0-0.0.master.20140130172954.git9257b30.el6.noarch ovirt-engine-setup-base-3.5.0-0.0.master.20140130172954.git9257b30.el6.noarch ovirt-engine-setup-plugin-ovirt-engine-3.5.0-0.0.master.20140130172954.git9257b30.el6.noarch ovirt-host-deploy-1.2.0-0.2.master.20140120.gitdeb0453.el6.noarch ovirt-host-deploy-java-1.2.0-0.2.master.20140120.gitdeb0453.el6.noarch ovirt-engine-setup-3.5.0-0.0.master.20140130172954.git9257b30.el6.noarch ovirt-iso-uploader-3.5.0-0.0.master.20140120.gitd18e6f7.el6.noarch ovirt-engine-websocket-proxy-3.4.0-0.7.beta2.el6.noarch ovirt-engine-webadmin-portal-3.4.0-0.7.beta2.el6.noarch ovirt-engine-userportal-3.4.0-0.7.beta2.el6.noarch ovirt-engine-3.4.0-0.7.beta2.el6.noarch ovirt-engine-sdk-python-3.5.0.0-1.20140128.gitf272277.el6.noarch ovirt-engine-setup-plugin-ovirt-engine-common-3.5.0-0.0.master.20140130172954.git9257b30.el6.noarch ovirt-engine-setup-plugin-websocket-proxy-3.5.0-0.0.master.20140130172954.git9257b30.el6.noarch ovirt-log-collector-3.5.0-0.0.master.20140130.git2122d59.el6.noarch ovirt-image-uploader-3.5.0-0.0.master.20140120.git7801707.el6.noarch ovirt-release-el6-10.0.1-3.noarch ovirt-engine-cli-3.4.0.3-1.el6.noarch ovirt-engine-dbscripts-3.4.0-0.7.beta2.el6.noarch ovirt-engine-backend-3.4.0-0.7.beta2.el6.noarch ovirt-engine-restapi-3.4.0-0.7.beta2.el6.noarch ovirt-engine-tools-3.4.0-0.7.beta2.el6.noarch
Jonas, could you specify your libvirt version and the output of `ifconfig <bridge-of-vnet26>` and that of `virsh -r dumpxml <name-of-your-vm>` ? Laine, shouldn't libvirt copy the MTU form the bridge to the tap device?
Sorry, I should have checked this, I did not know the architecture SAN Link encap:Ethernet HWaddr D0:67:E5:F9:2E:1C inet addr:192.168.43.11 Bcast:192.168.43.255 Mask:255.255.255.0 inet6 addr: fe80::d267:e5ff:fef9:2e1c/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:386710758 errors:0 dropped:0 overruns:0 frame:0 TX packets:384811517 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes:2343055570732 (2.1 TiB) TX bytes:396696696938 (369.4 GiB) So I guess what seem to have happened here is that the engine-settings on this bridge is somehow lost ?
Created attachment 869710 [details] San net config
SAN 8000.d067e5f92e1c no em2 vnet0 vnet11 vnet13 vnet16 vnet19 vnet21 vnet26 vnet4 vnet6 vnet8 vnet9
Not allowed to raise the MTU beyond 1500 on the bridge. [root@ft admin]# ifconfig SAN mtu 9000 SIOCSIFMTU: Invalid argument
Have you changed the MTU while the network has already been in use by VMs? Does the issue reproduce on new hosts or after boot? Could you share your ifcfg-SAN, and supervdsm.log with the setupNetworks() setting it?
>Have you changed the MTU while the network has already been in use by VMs? No I have after installing this system not changed anything related to the network config. > Does the issue reproduce on new hosts or after boot? I currently only have a single host and it can not easily be rebooted. I'll do my best trying to make time for maintenance and inform my users. > Could you share your ifcfg-SAN, [root@ft network-scripts]# more ifcfg-SAN # Generated by VDSM version 4.14.1-2.el6 DEVICE=SAN ONBOOT=yes TYPE=Bridge DELAY=0 IPADDR=192.168.43.11 NETMASK=255.255.255.0 BOOTPROTO=none MTU=9000 NM_CONTROLLED=no STP=no > and supervdsm.log with the setupNetworks() setting it? Attached
Created attachment 869801 [details] The supervdsm log
I see no apparent clue on why the bridge ignored the MTU setting. I'd appreciate a reproduction - maybe with another network on the same host. Would an explicit ifconfig vnet0 mtu 9000 for each vnet* would make it possible to set the mtu on the bridge and provide an immediate workwaround?
> I see no apparent clue on why the bridge ignored the MTU setting. I'd > appreciate a reproduction - maybe with another network on the same host. I do have a free NIC in this host that I could try and play around with.. >Would an explicit > ifconfig vnet0 mtu 9000 >for each vnet* would make it possible to set the mtu on the bridge and provide >an immediate workwaround? I did try this yesterday, it made no difference. I have now rebooted the host and the bridge now is back to MTU 9000 SAN Link encap:Ethernet HWaddr D0:67:E5:F9:2E:1C inet addr:192.168.43.11 Bcast:192.168.43.255 Mask:255.255.255.0 inet6 addr: fe80::d267:e5ff:fef9:2e1c/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:9000 Metric:1 RX packets:28617 errors:0 dropped:0 overruns:0 frame:0 TX packets:21750 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes:191312618 (182.4 MiB) TX bytes:5328456 (5.0 MiB) I fired up my first VM and that too now get the right MTU vnet0 Link encap:Ethernet HWaddr FE:1A:4A:2F:D2:A6 inet6 addr: fe80::fc1a:4aff:fe2f:d2a6/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:9000 Metric:1 RX packets:8 errors:0 dropped:0 overruns:0 frame:0 TX packets:9 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:500 RX bytes:648 (648.0 b) TX bytes:648 (648.0 b) I'll do my best to try and reproduce this again.
Please reopen this bug when it reproduces or substantial new information is found.