Tom,  do you know yet whether this problem only affects a small number of users and/or is likely hardware/driver related (nic etc)?

 

I have never had this network-dropped problem, and it is my impression that most unRAID users have not had it either, or would have reported it.  I have a Marvell networking chipset, quite common currently, and use the forcedeth driver.  Many others use Intel Pro family NIC's with the e1000 driver, without a problem.  I've seen the r1000 and r8169 drivers mentioned elsewhere for Realtek chipsets.

 

unRAID has to rely on the quality of others' Linux drivers for the various hardware components we use.  So, although it could be a problem with the unRAID system, it could also be a bad driver, or faulty or incompatible hardware.

 

It would be useful if users with this problem would report their hardware, as trueimage did above.  Possibly more helpful would be to report the chipsets and the Linux drivers used.  You can find this out by examining your syslog and searching eth0, or using the lsmod command at a console or Telnet prompt.  Perhaps this way, we can determine if there are common drivers or chipsets involved.

 

I noticed that Corys8646's syslog indicates using the networking driver sky2, which I have only seen in one other syslog.  Because it is less commonly used, perhaps it is less well tested?  When Corys8646 switched to a different NIC, there was probably a switch to a different networking driver too, perhaps a better one.  I only mention this as an example, of a change in hardware also causing a change in software.

 

New user here.

 

Just build my server, P5LD2 with onboard nic.  Was working fine until I attempted to copy a fileset larger than 4.5 G.  Same as above poster's have mentioned, network died.  Came back on re-boot.

 

Have not yet tried the fix mentioned above (will do shortly and report back).  Hopefully this is not a showstopper as I am still evalutating wether to upgrade to the paid version and add more drives.  Need the ability to move large file sets like HD images, etc. reliably

 

Cheers

 

Dave

FWIW, after applying the nolapic argument suggested above I was able to successfully copy a 12 gig file set to my unraid box without the network collapsing.

 

Will be watching it for a couple of days to assess stability.

 

Dave

Please try the following to see if it solves the 'dropped network' problem.

 

Edit the file 'syslinux.cfg' on the Flash device.  On the line that reads:

 

append initrd=bzroot rootdelay=10

change to

 

append initrd=bzroot rootdelay=10 nolapic

 

Then reboot your server.

 

I've applied this a few days ago.

 

I've been copying a few 6-9 GB files one at a time with no problems.

 

Now, I'm trying several files totalling 50GB.

 

Everything seems ok since applying this fix.

APIC support has always been problematic in linux - at least that's our experience - and from counting the number of problems reported on various mailing lists which have been solved by turning off APIC support, seems to be the experience of others as well.

 

We had APIC support turned off in unRAID pre-4.0 releases (based on 2.4 kernel) for exactly this reason.  When we migrated to linux 2.6 kernel, we re-enabled APIC and the problems we used to see no longer happened, so I figured, ok the kernel guys have finally hardened this stuff.  But apparently there are still problems with APIC.

 

APIC is really only useful in a multi-CPU environment anyway, e.g., dual-core.  So we'll turn it off in the next release and re-address this issue in the future when it might be useful to have dual-core CPU's.

 

Thanks for the post that this helped so far, i'll have to pull my server down to try this.  I'll post results at that time.

 

 

Tom, more info for you that i did not share earlier.

 

This issue occurs when i have heavy file movement and eventually loose my network connection to the server completely.

Rebooting brings everything back up.  the web management shows no errors.

 

I have 512MB of RAM and before the other memory post came into the picture, i've ordered more RAM so i can run at 1GB just in case.

I do not write files in user share if you're wondering.

 

I have noticed when the initial issue takes place where i need to reach the unraid via IP and not it's name, that the memory usage is peaking it's max 512.  This was the reason for me to get more RAM, if there is a leak that maybe attributing to this issue then i suppose I'm only buying a little bit more time but the issue will eventually happen again.

 

Also i can replicate this issue in couple of ways.

 

If I copy multiple large (1GB+) files.  Lets say 4 of those being copied at the same time, I end up with this issue at times.

 

or

 

If I copy multiple sessions in various files of lets say 12 or greater writes at the same time then I end up with this issue eventually.

Both events above takes time before this issue occurs and it's never instant...

 

I originally mapped my unRAID folder as the download folder for a newsgroup software. 

I started running into issues after that as this software writes/builds files from 12 to 15 sessions depending on how many different nzb files i'm pulling down.

 

Anyway, I hope this helps a bit as the two areas above can be used to replicate the issues.  I plan on giving the 'nolapic' a try once i pull the server down in the days to come. 

 

Thanks,

M

 

 

 

I was playing movies with guests off the server and twice I lost network connection.

Was playing two divx files, so the network traffic was very limited.

As reported, network goes down, but the unraid server is up.

Rebooting the server brings the network connection back... for a while.

I have run the nolapic fix, we will see how it works...

Please try the following to see if it solves the 'dropped network' problem.

 

Edit the file 'syslinux.cfg' on the Flash device.  On the line that reads:

 

append initrd=bzroot rootdelay=10

change to

 

append initrd=bzroot rootdelay=10 nolapic

 

Then reboot your server.

 

I've applied this a few days ago.

 

I've been copying a few 6-9 GB files one at a time with no problems.

 

Now, I'm trying several files totalling 50GB.

 

Everything seems ok since applying this fix.

 

 

Just an update - everything seems to be working. no network drops at all so far, xferred like 100 GB in the past few days.

Seems to have fixed the issue for me as well.

 

"Re-upgraded" back to 4.1 from 3.0 with the nolapic fix and I've been transferring with no issues for the last couple of days.

 

 

Still connected, so far so good for me too... I haven't transferred any data, but I will this weekend.

Please try the following to see if it solves the 'dropped network' problem.

 

Edit the file 'syslinux.cfg' on the Flash device.  On the line that reads:

 

append initrd=bzroot rootdelay=10

change to

 

append initrd=bzroot rootdelay=10 nolapic

 

Then reboot your server.

 

 

 

This worked for me... it's not loosing network connection any more.

 

Thanks Tom

The only time I have been able to reproduce something like this problem, is when testing the s/w by introducing massive numbers of disk errors - there is a memory leak we've tracked down, but I don't think this is causing what most people are posting about in this thread.

 

If system becomes unreachable via network, the way to get the system log is:

1. log in via console (keyboard/monitor)

2. copy the system log to the flash:

  cp  /var/log/syslog  /boot/syslog.txt

3. reboot system

4. send the syslog.txt file to me (it will be in the Flash share).

 

Ultimately the source of this is probably hardware - but not necessarily the NIC, could be a side-effect of something else failing  :'(

 

I have lost connection again yesterday night, will try this time to get a syslog copy on the flash before I reboot and copy it here.

Rgds

I have attached the syslog

 

Very annoying, have had friends over today and the server lost the network 3 times whilst watching the movie.

Had to go to the server and reboot it three times. Anyone can advise on a network compatible card?

I am ready to try anything to be honest with you....

I have attached the syslog

 

Very annoying, have had friends over today and the server lost the network 3 times whilst watching the movie.

Had to go to the server and reboot it three times. Anyone can advise on a network compatible card?

I am ready to try anything to be honest with you....

 

Won't allow me to attach the syslog.... will try later on....

It has happened again and again. Basically whenever I play a movie over the network, and I do it only once a week, I loose connection, whilst the server is running fine. The only way around is to phisically go to the server and type reboot.

I have had enough of this I tell you. I will buy a network card to see if this solves the problem. Any hardware suggestions?

Run the syslog, still cannot attach it. Is it just me, or you cannot attach any text files or zips?

choose a NIC listed on the WIKI as working.

/Rene

I have ordered this NIC as suggested by mr Limetech himself.

 

Trendnet TEG-PCI TXR

 

http://trendnetdirect.co.uk/index.php?mod=product&id_prd=100

 

Should arrive on Monday. Will install and report if the problem still occours.

well well well.

Received the PCI network card, opened my case, and realised that the mboard has only to PCI slots.

 

Mboard I use is the Intel Desktop microATX Motherboard D915GAGLK.

 

Expansion Slots Total (Free)

 

1 processor - LGA775 Socket

4 memory ( 2.5 V ) - DIMM 184-pin

1 PCI Express x16

1 PCI Express x1

2 PCI

 

This means my 2 PCI slots are already used by the expansion cards for the Sata drives....

 

Can anyone suggest a PCI Express network card that works with Unraid then?

 

Thanks for your help, the search continues....

 

There you go, lost the bloody network again...

 

Syslog attached

 

 

there seems to be a clue in the last few lines, has something to do with DHCP

 

...

"Mar 17 00:07:22 Tower emhttp[1454]: shcmd (86): killall -HUP smbd

Mar 17 11:18:49 Tower dhcpcd[1325]: sending DHCP_REQUEST for 192.168.1.70 to 192.168.1.254

Mar 17 11:18:49 Tower dhcpcd[1325]: dhcpIPaddrLeaseTime=86400 in DHCP server response.

Mar 17 11:18:49 Tower dhcpcd[1325]: dhcpT1value is missing in DHCP server response. Assuming 43200 sec

Mar 17 11:18:49 Tower dhcpcd[1325]: dhcpT2value is missing in DHCP server response. Assuming 75600 sec

Mar 17 11:18:49 Tower dhcpcd[1325]: DHCP_ACK received from  (192.168.1.254)

Mar 17 23:18:49 Tower dhcpcd[1325]: sending DHCP_REQUEST for 192.168.1.70 to 192.168.1.254

Mar 18 01:36:33 Tower dhcpcd[1325]: sendto: No buffer space available

Mar 18 01:36:33 Tower dhcpcd[1325]: broadcasting DHCP_REQUEST for 192.168.1.70

Mar 18 01:36:33 Tower dhcpcd[1325]: sendto: No buffer space available

Mar 18 01:36:33 Tower kernel: [181205.456520] sky2 eth0: disabling interface

Mar 18 01:36:33 Tower kernel: [181205.457723] sky2 eth0: enabling interface

Mar 18 01:36:33 Tower kernel: [181205.459549] sky2 eth0: ram buffer 48K

Mar 18 01:36:33 Tower dhcpcd[1325]: broadcasting DHCP_DISCOVER

Mar 18 01:36:34 Tower ifplugd(eth0)[1413]: Link beat lost.

Mar 18 01:36:36 Tower kernel: [181208.642729] sky2 eth0: Link is up at 1000 Mbps, full duplex, flow control both

Mar 18 01:36:37 Tower ifplugd(eth0)[1413]: Link beat detected.

Mar 18 04:31:42 Tower dhcpcd[1325]: sendto: No buffer space available

Mar 18 04:31:42 Tower kernel: [191709.827103] sky2 eth0: disabling interface

Mar 18 04:31:43 Tower ifplugd(eth0)[1413]: Link beat lost.

Mar 18 04:31:43 Tower kernel: [191710.546750] sky2 eth0: enabling interface

Mar 18 04:31:43 Tower kernel: [191710.548575] sky2 eth0: ram buffer 48K

Mar 18 04:31:46 Tower kernel: [191712.983280] sky2 eth0: Link is up at 1000 Mbps, full duplex, flow control both

Mar 18 04:31:46 Tower ifplugd(eth0)[1413]: Link beat detected."

...

 

No buffer space available?

 

A guy that has exactly the same problem I experience.

Never used to happen, then all of a sudden no network connection, nothing wrong with the server, reboot is the only way to fix it...

 

http://archive.linuxvirtualserver.org/html/lvs-users/2002-09/msg00243.html

"Link beat lost" messages would seem to indicate an issue with the cabling to the router, the network card, or the router. 

 

I would start with checking the lan cable to ensure it is seated well, then if the problems continue,

replacing the lan cable with a new one of known high quality,

then trying a different port on the router...

 

I already swapped cables, and ports on the router.

Currently using the network connection of my motherboard.

I just realised that I cannot use a pci network card as I don't have any more PCI connections free, only PCI-X.

gnollo-

Please type this command from telnet session and post the output:

 

df

 

Thanks.

 

Tower login: root

 

[Disconnect bypassed -- root login allowed.]

Linux 2.6.22.5.

root@Tower:~# df

Filesystem          1K-blocks      Used Available Use% Mounted on

/dev/sde1              501280    30752    470528  7% /boot

/dev/md1            488371640 478044588  10327052  98% /mnt/disk1

/dev/md10            488371640 156852996 331518644  33% /mnt/disk10

/dev/md3            488371640 479191436  9180204  99% /mnt/disk3

/dev/md2            488371640 463643560  24728080  95% /mnt/disk2

/dev/md5            488371640 482961184  5410456  99% /mnt/disk5

/dev/md9            488371640 475773740  12597900  98% /mnt/disk9

/dev/md4            488371640 457835496  30536144  94% /mnt/disk4

/dev/md6            488371640 483668660  4702980 100% /mnt/disk6

/dev/md8            488371640 485426156  2945484 100% /mnt/disk8

/dev/md7            488371640 485896084  2475556 100% /mnt/disk7

shfs                4883716400 4449293900 434422500  92% /mnt/user

root@Tower:~#