April 3, 201412 yr Is any one else seeing this in their Xen Linux VMs? This is my vif line... vif = [ 'mac=00:16:3e:d3:e5:c4,bridge=br0,model=e1000' ] Thank you
April 13, 201412 yr I just saw the same message in my ArchVM. I'm using the appliance created by ironicbadger. I did a quick Google search on the message and it appears that it may mean when you get that message your suffering from packet loss. I had also just noticed that my internet connection on the VM was being flaky. Anyone else seeing this or know how to fix it? Thanks, Doug
April 17, 201412 yr This command fixed it for me: ethtool -K eth0 tso off gso off Please someone step in and tell me if this is doing more harm than good.
April 17, 201412 yr This command fixed it for me: ethtool -K eth0 tso off gso off Please someone step in and tell me if this is doing more harm than good. It stopped the rocket rides for me. I also added gro off. The only way to see if it does more harm than good is to test your vm's network connection with on or off and check for packet loss in the vm. Only thing is it will not be permanent. A reboot of the vm will turn offload back on. You need to add a script to execute the command when eth0 is up. Also ethtool -k eth0 will show you whats turned on.
April 22, 201412 yr This command fixed it for me: ethtool -K eth0 tso off gso off Please someone step in and tell me if this is doing more harm than good. I tried this command inside the VM but that command is not found. I was able to run it outside the VMs, but it doesn't seem to have any impact. What else can I try? My console is constantly scrolling now with this rocket message. Doug
April 23, 201412 yr +1 I did install ethtool in the arch vm, and have turned off tso, gro, gso as of now. No rocket rides for the past three minutes, so that's something! I shall report back after a decent test interval... EDIT: Oh, for Doug: pacman -S ethtool will install that for you in the VM, and you can try along with me!
April 23, 201412 yr Ok. So don't do this. Just a few minutes in, my server crashed. Hard. I have pretty much narrowed it down to using these options with sabnzbd running, as it was causing most of the rocket rides in the first place, at least as near as I can tell. It's getting late, and this is my production server, so further testing (from me anyway) will probably have to wait. It crashes with some kind of CPU rcu_sched error, I tried to leave a telnet window and a dmesg window open, but it crashed without writing anything except to the monitor I had hooked up.
April 23, 201412 yr Im seeing both the rocket rides and CPU rcu_sched error, however I havent installed ethtool nor have I done anything other than install the usenet programs and webmin on that vm. @ Pengrus, do you have snap installed? The reason I ask is because after installing snap was when I started to get the cpu errors (every 4-6 hrs for 3 days) uninstalled snap last night and so far Im 18 hrs in with no cpu errors. (Still riding that damn rocket though) Sent from my SCH-I605 using Tapatalk
April 23, 201412 yr Nope, no SNAP. I run powerdown/apcupsd on the server, MySQL, sab/sick/cp on an arch vm, and a mail server (wip) on a ubuntu server vm. Maybe I'm overcommitting resources somehow, I only have 8gb of ram. Stopping SAB stopped the rocket rides for the most part, though it did end up crashing my mysql instance a few times, not really sure how. My swag is that the rocket rides essentially DOSed the network in the house, thus preventing nfs and the SQL socket from communicating. Or something, I have no damn idea. Kind of defeats the purpose of having the VMs though, if they are actually more of a headache than running the plugins directly on the server!
April 24, 201412 yr So like I said before, no SNAP, but I did use ironic's archvm image and mounted my cache drive via nfs for storage space for sabnzbd. It appears to be this interface that is dropping all the packets. When I umount the cache drive and restart sab, no problems. Unless you count having 5G of space a problem! deionmann, are you mounting your drive via nfs?
April 24, 201412 yr I'm mounting both from a cache only share, and an array share. and yes both are connected to Sabnzbd basically on this vm I have Sabnzbd,Deluge,Transmission,Sickbeard(TPB Fork), couchpotato, headphones, and webmin all running on Ironic's ArvcVM w/ 4G of ram additionally I have an ArchVM running with 2G and plex and a HVM running untangled with one nic passed through for wan, and the lan is a vif on the main bridge (named lan1 on my system) here's my .cfg for the sabnzbd vm (the only vm riding the rocket) # Created for the unRaid community by IronicBadger. # Last updated 16/04/2014. Enjoy! name = "dom3" # change me as required bootloader = "pygrub" memory = 4096 # update me to 4096 if using sabnzbd vcpus = '2' # update me to no of CPUs allocated disk = [ 'phy:/mnt/xendisk2/ds2/dom3/dom3.img,xvda,w', # 'phy:/mnt/user/Software/data.img,xvdb,w' ] # replace xx:xx:xx with random a1:b2:c3 mac address vif = [ 'mac=00:16:3e:c1:c1:d1,vifname=Cthulhu,bridge=lan1' ] and here's my fstab entry # # /etc/fstab: static file system information # # <file system> <dir> <type> <options> <dump> <pass> # /dev/xvda1 UUID=93ec2c22-36c1-487c-a888-adde602a16fe / ext4 rw,relatime,data=ordered 0 1 #Godzilla array mount //10.0.0.200/media /mnt/media cifs noauto,x-systemd.automount,rw,guest,file_mode=077,dir_mode=077 #Godzilla cache only mount //10.0.0.200/process /mnt/process cifs noauto,x-systemd.automount,rw,guest,file_mode=077,dir_mode=077 BTW I'm pretty sure for me, the CPU errors were definitely caused by SNAP since I am still running cpu error free since uninstalling it
April 29, 201412 yr With some further testing, I have narrowed the errors down to some way the VMs are interfacing with the cache drive. Any time I try to get data off the VM and to the cache/array, the rocket rides start. Several times a second. This happens with NFS mounts and SMB/CIFS mounts. It also happens with FTP. I have not tried AFP, though I don't even know how I would do that. I am in the middle of fixing the mySQL plugin to work natively on unraid and abandoning the VM for that specific purpose, as every time XBMC tries to write watched status or whatever back to the database, it runs the risk of failing the write. When it fails the write, the database disappears, and I have to restart mysql to get it back. Does anyone know why this happens? Google has illuminated me exactly zero on the subject. I have always had a large number of dropped transmission packets on that NIC, it's the Intel one on the Supermicro board, any way to fix that maybe? Just shooting in the dark at this point!
May 6, 201412 yr Im an idiot. SNAP is gone, but the cpu errors are back. Still riding the rocket, going to uninstall sab on my vm and see if that helps Sent from my SCH-I605 using Tapatalk
May 19, 201412 yr I'm mounting both from a cache only share, and an array share. and yes both are connected to Sabnzbd basically on this vm I have Sabnzbd,Deluge,Transmission,Sickbeard(TPB Fork), couchpotato, headphones, and webmin all running on Ironic's ArvcVM w/ 4G of ram additionally I have an ArchVM running with 2G and plex and a HVM running untangled with one nic passed through for wan, and the lan is a vif on the main bridge (named lan1 on my system) here's my .cfg for the sabnzbd vm (the only vm riding the rocket) # Created for the unRaid community by IronicBadger. # Last updated 16/04/2014. Enjoy! name = "dom3" # change me as required bootloader = "pygrub" memory = 4096 # update me to 4096 if using sabnzbd vcpus = '2' # update me to no of CPUs allocated disk = [ 'phy:/mnt/xendisk2/ds2/dom3/dom3.img,xvda,w', # 'phy:/mnt/user/Software/data.img,xvdb,w' ] # replace xx:xx:xx with random a1:b2:c3 mac address vif = [ 'mac=00:16:3e:c1:c1:d1,vifname=Cthulhu,bridge=lan1' ] and here's my fstab entry # # /etc/fstab: static file system information # # <file system> <dir> <type> <options> <dump> <pass> # /dev/xvda1 UUID=93ec2c22-36c1-487c-a888-adde602a16fe / ext4 rw,relatime,data=ordered 0 1 #Godzilla array mount //10.0.0.200/media /mnt/media cifs noauto,x-systemd.automount,rw,guest,file_mode=077,dir_mode=077 #Godzilla cache only mount //10.0.0.200/process /mnt/process cifs noauto,x-systemd.automount,rw,guest,file_mode=077,dir_mode=077 BTW I'm pretty sure for me, the CPU errors were definitely caused by SNAP since I am still running cpu error free since uninstalling it Why are you using "noauto"?
May 19, 201412 yr It was my solution to the DHCP delay, im running transmission on that vm and I found it would error out everytime with normal cifs. the noauto switch fixed the issue for me. (For some reason or another autofs didnt work well for me - probably because im a linux noob xP) Sent from my SCH-I605 using Tapatalk
May 19, 201412 yr hmmm so right after boot (without noauto) the mounts weren't automounting on first access? That is the behavior I am seeing. But I'm also not as fast as a service and Plex fails very gracefully when resources are not immediately available. It just retries in a few moments. Probably because it was written to assume removable media might come and go. In any case ... so did you try using a static IP? or at least reserving the IP in your router? That didn't help? Im' just trying to scope out when and where x-systemd.automount is failing because for me it 100% solved the problem of shares not mounting at all. But like I said I can't speak to any race conditions. Did you see PeterB's post http://lime-technology.com/forum/index.php?topic=31687.msg300532#msg300532 to see about delaying transmission until the network was up? PS if you are on 6.0b5a, did you turn off STP in unraid network setup?
May 19, 201412 yr hmmm so right after boot (without noauto) the mounts weren't automounting on first access? ... did you try using a static IP? or at least reserving the IP in your router? That didn't help? I have a static ip setup on arch through netctl. They werent mounting on first access before I setup noauto, after noauto mounting works like a charm every time. Cant tell you if the static ip it helped honestly because setting hostname/static ip/changing root psswd/and installing webmin are the first things I do before installing any other program Did you see PeterB's post http://lime-technology.com/forum/index.php?topic=31687.msg300532#msg300532 to see about delaying transmission until the network was up? PS if you are on 6.0b5a, did you turn off STP in unraid network setup? I saw the post, but that was after I solved my issues with noauto, think It might solve the rocket rides if I were to use that technique instead of noauto? I am running 6.0Ba, And no, I didnt try turning off STP because my mobo has 4 nics that I use for my lan bridge ( https://www.google.com/url?sa=t&source=web&rct=j&ei=zCh6U8DlJLTQsQT0_oGwDg&url=http://www.newegg.com/Product/Product.aspx%3FItem%3DN82E16813131723&cd=3&ved=0CCsQFjAC&usg=AFQjCNE69oen0jzrin_xAJR0rVJlpPX6bQ&sig2=4JmgHH3K3Nj4pQLPADat5A ) . My understanding is that turning off STP doesnt work well with multiple nics in the same system? Sent from my SCH-I605 using Tapatalk
May 19, 201412 yr hmmm so right after boot (without noauto) the mounts weren't automounting on first access? ... did you try using a static IP? or at least reserving the IP in your router? That didn't help? I have a static ip setup on arch through netctl. They werent mounting on first access before I setup noauto, after noauto mounting works like a charm every time. Cant tell you if the static ip it helped honestly because setting hostname/static ip/changing root psswd/and installing webmin are the first things I do before installing any other program So you had to ADD noauto to get automounting to work? Weird. Or do you mean you had to add it to get it to mount at all? Do you need to manually issue mount -a? If you are so inclined it would be interesting to see what happened if you added "auto" as opposed to not having auto/noauto at all. I wouldn't expect the static IP to help with automounting itself. Just with any delays in the network coming up. Did you see PeterB's post http://lime-technology.com/forum/index.php?topic=31687.msg300532#msg300532 to see about delaying transmission until the network was up? PS if you are on 6.0b5a, did you turn off STP in unraid network setup? I saw the post, but that was after I solved my issues with noauto, think It might solve the rocket rides if I were to use that technique instead of noauto? I am running 6.0Ba, And no, I didnt try turning off STP because my mobo has 4 nics that I use for my lan bridge ( https://www.google.com/url?sa=t&source=web&rct=j&ei=zCh6U8DlJLTQsQT0_oGwDg&url=http://www.newegg.com/Product/Product.aspx%3FItem%3DN82E16813131723&cd=3&ved=0CCsQFjAC&usg=AFQjCNE69oen0jzrin_xAJR0rVJlpPX6bQ&sig2=4JmgHH3K3Nj4pQLPADat5A ) . My understanding is that turning off STP doesnt work well with multiple nics in the same system? Yes indeed you might need STP if you have multiple nics ... but I thought that was more of an issue if you were briding them together or something like that. Not just assigning them to individual VMs. Basically if you are creating any loops in your network. http://lime-technology.com/forum/index.php?topic=32682.msg300245#msg300245 I'm hardly an expert of course.
May 19, 201412 yr So you had to ADD noauto to get automounting to work? Weird. Or do you mean you had to add it to get it to mount at all? Do you need to manually issue mount -a? If you are so inclined it would be interesting to see what happened if you added "auto" as opposed to not having auto/noauto at all. I wouldn't expect the static IP to help with automounting itself. Just with any delays in the network coming up. I just had to add it to get automount to work. Worked fine issuing a mount -a command Ill change the noauto to auto and post my outcome Sent from my SCH-I605 using Tapatalk
May 19, 201412 yr So you had to ADD noauto to get automounting to work? Weird. Or do you mean you had to add it to get it to mount at all? Do you need to manually issue mount -a? If you are so inclined it would be interesting to see what happened if you added "auto" as opposed to not having auto/noauto at all. I wouldn't expect the static IP to help with automounting itself. Just with any delays in the network coming up. I just had to add it to get automount to work. Worked fine issuing a mount -a command Ill change the noauto to auto and post my outcome Sent from my SCH-I605 using Tapatalk that is just weird though isn't it?
June 1, 201412 yr Just to get back on topic, I'd like to chime in that I'm also getting these "skb rides the rocket" alerts. For me they are most prevalent when SAB is working, as that's when the majority of the data transfer is taking place. I do notice slowdowns on the transfer speeds and download rates when these errors occur.
June 1, 201412 yr Theory of cause of this issue ( from another thread): http://lime-technology.com/forum/index.php?topic=31687.msg309363#msg309363
June 1, 201412 yr Theory of cause of this issue ( from another thread): http://lime-technology.com/forum/index.php?topic=31687.msg309363#msg309363 Not likely, my mounts are all manually added via fstab, so I don't use /net/ at all. My research so far leads me to believe this is an issue with how Xen handles fragmented packets. At first I thought it was a relation to the MTU and having scatter gather enabled, but setting my MTU to 1500 and disabling sg didn't give me positive results. More likely this is a Xen kernel bug, as some dev chatter have narrowed some down possible causes for the extra frag pages, but it doesn't sound easily fixed. I'm hoping to see results in a future Xen update, because the performance hit with this error is huge.
June 20, 201412 yr Anyone know if this is fixed in beta6? Can't seem to get mysql working "natively" with the available 64-bit unraid plugins...WAF currently plummeting!! Would very much like to offload this and other functions to some nice and tiny arch vms...
Archived
This topic is now archived and is closed to further replies.