I went to check on my server this morning and turn on a movie for my son, and lo and behold, the server is no longer accessible via the web interface.  This had happened last week, the server asked for an IP address and none was assigned, but a restart fixed it.  No such luck this time around.  It has been down all day, nothing I try is helping.  The machine is a X8SIL-F-O board with onboard dual NIC and IPMI, a bunch of EARS drives and a few 2.5" laptop drives up top.  One of those failed a few days ago, but there was nothing on it, so I just rebuilt the array without that drive.

 

Funny thing is, the IPMI interface still works while the server is operating if I use the dedicated connection.  it ALSO works if the server is off, but at some point in the boot sequence, eth0 stops functioning at all.  No IPMI and no network.  I had the DHCP lease permanently assigned, so I tried not doing that, and assigning a static IP outside of the DHCP pool, but nothing works.  I'll be honest, I am beginning to get a touch frustrated with this problem, as network issues have been the hallmark of this build.  The phrase "timed out while awaiting a response from the DHCP server" gives me chills.

 

Also, if it matters, I have a router behind a router situation here, where I use the AT&T RG to connect to the internet, then my linksys e3000 is connected in the DMZ, and all devices connect to it.  Could that be a problem?

 

I tend to think that it is something in the unraid slackware distro or something, although I know others use this board with no issues.  I say that because the IPMI interface works when the server is off, but not on.  Weird.

 

Thanks in advance, please let me know what you guys think!

Syslogs.zip

I am still going through your syslog, but I picked up the following:

Mar 24 17:43:40 Tower kernel: 0000:04:00.0: eth1: (PCI Express:2.5GB/s:Width x1) 00:25:90:33:5d:eb

Mar 24 17:43:40 Tower kernel: 0000:04:00.0: eth1: Intel® PRO/1000 Network Connection

Mar 24 17:43:40 Tower kernel: 0000:04:00.0: eth1: MAC: 3, PHY: 8, PBA No: 0101ff-0ff

 

I have this same board and looked at my syslog to ensure my IPMI was not picked up as eth1. It was not, I have mine disabled in the bios. Is it possible in your troubleshooting you hooked your lan cable to eth1? unRAID will only see eth0 if I am not mistaking.

Mar 24 17:43:40 Tower ifplugd(eth0)[1603]: Initialization complete, link beat not detected.

 

http://linux.die.net/man/8/ifplugd this page discribes what the ifplugd daemon does. Could be that it is connected to the wrong port.

 

Alright, I should have waited until I got through the entire syslog. All the above seems to be working as I found a link beat a little further down. Your problem seems to have started here:

 

Mar 28 06:13:01 Tower kernel: e1000e: eth0 NIC Link is Down

Mar 28 06:13:01 Tower ifplugd(eth0)[1603]: Link beat lost.

Mar 28 06:13:04 Tower kernel: e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX

Mar 28 06:13:04 Tower ifplugd(eth0)[1603]: Link beat detected.

Mar 28 07:58:32 Tower kernel: e1000e: eth0 NIC Link is Down

Mar 28 07:58:33 Tower ifplugd(eth0)[1603]: Link beat lost.

Mar 28 07:58:35 Tower kernel: e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX

Mar 28 07:58:35 Tower ifplugd(eth0)[1603]: Link beat detected.

 

I would take a look at your lan cables first, if that is not it then start taking out a component at a time to find the issue. It could be your server nic, but I would suspect that last.

 

I have tried several different cables, both 5e and cat 6.  No luck.  As far as removing components, I'm really not sure what else I can remove.  I will try disconnecting the router from the Uverse gateway when I get home, but I don't know that that is the issue, given that the IPMI works on the dedicated interface all the time, and eth0's interface when the server is powered down.  IPMI can get an IP address, why can't I?  Although, XBMC (on ubuntu 10.04) has a problem accessing the internet for like weather and stuff.  It takes several clicks on the widget to get it to come up.  Maybe this is a gateway assignment problem or something, I dunno.  I need to go to LAN school apparently.

 

This makes no sense to me.  When I first built the server, it didn't like the cable I used.  No lights at all, except on the IPMI interface.  I thought fine, must be a GigE thing, since the IPMI is 10/100.  Changed the cable, and it worked for a day until it would drop out, ostensibly when asking for a new IP lease.  I figured it was a poorly configured Windows PC on the network that was acting as if it was the DHCP server, but not handing out any addresses.  So I disabled that, and everything worked fine.  For several weeks it worked fine until I went out of town and it dropped out in the middle of my wife watching a movie.  Score one for the "don't cut the cable cord" team in my house.

 

To reiterate, nothing has changed.  I hadn't made any changes, so I rebooted the server, and everything worked for a few days until yesterday morning.  The windows PC isn't even on right now. 

 

It is either not successfully talking TO the DHCP server on 192.168.2.1 (anymore) or in hearing FROM the router.

 

ethtool sees the interface fine, and as you can see from the syslog, it sees it and detects the link beat or whatever, and then it drops out.  If anyone knows a way to change to eth1 maybe or disable the ports in the bios, I'm open to suggestions.

 

I have a netgear g311 nic being delivered today, but this is a brand new X8SIL-F-O that others use with no issue. 

 

Again, any suggestions are welcomed, especially if any of you with this board or using it with Uverse have had to make any changes.  Thanks again.

I feel your frustration, especially with the wife and cable thing. I am fighting that myself. I too am not an expert with networking. I read and tweak, have been able to keep mine up, but mine is simple modem, router, switch, then computers. There is a setting in the bios that will allow you to disable eth0 and enable eth1. It will come up eth0 in the syslog though. I have a week of testing that I am doing with the fan speed script before I will reboot my server, trying to get to 30 days with different temps and what not to ensure it works for me.

 

I would take as much out of the equation as possible, perhaps moden, router, server or router, server and see if it picks up and keeps the connnection.

I'd suspect the linksys e3000 if two machines have had trouble.

Thanks man!  Yeah, who knew wives don't like being beta testers?  I tried unplugging the linksys router from the gateway to try using just as a switch, but I think I jacked setting that up.  I tried resetting the subnets, reassigning the IP addresses, anything I could think of, but no dice.  I tried plugging it into the Uverse gateway, and like magic it returns.  I'm leaving it connected all tonight and tomorrow to see if it drops out when the address renews, but it seems that the router just won't play with the server.  It's pretty ridiculous.  This is a linksys e3000 running both the stock firmware or dd-wrt.  dd-wtf more like.

 

But anyway, I'm pretty sure I need GigE for streaming HD, so maybe I can pick up a solid switch or start trying to fix the E3000 to just be a switch.  Given this issue and some of the funny happenings with the HTPC, perhaps this router is just hosed somehow.

 

Thanks again for checking back on me (i want that cord cut!), I'll make sure and post tomorrow with an update.

@dgaschk:  it appears you are correct!  I hope it stays that way.  At this point just KNOWING what the problem component is would be a huge improvement.  Can I RMA a router do you think?  Especially since I've had it for several months...

 

Thanks for the help, I'm not going to jinx myself my solving the thread just yet, but things are looking a little better...

Why not run Static IP opposed to a Dynamic IP. I had a switch a while back that would for some odd reason just shut off, but when it returned I always had an IP address and connectivity. Of course it took a bit longer to trouble shoot and figure out what was failing and why I would loose a connection and then all of the sudden be back on, but I never lost a connection for very long. LOL

 

Also if you have issues on two machines I'd bet its the router too.

Make sure you do not have any rogue devices plugged into your router/switch.  I had this happen recently to me.  I have 2 different cable modems...one for internet and one for phone.  The phone modem has an RJ45 jack so I figured "why not plug this into teh network too...maybe they use to upgarde firmware or whatever".

 

BAD IDEA!!!

 

My entire network was up and down constantly.  The thing was continually ARP'ing (if that is teh right word).  The more I thought about it the mroe it made sense.  That port was probably designed more to act like a router/dhcp server/gateway...whatever...more so than a client.  Once unplugged, everthing settled.

 

Yeah...I'm that way.  Plug it in a see what happens.    Funny thing is I now remember having teh same issue with the etehrnet port on my DVR.  Not sure what the issue was there but teh symptoms were the same.

 

Just a thought.

 

John

Reading your original post, a little warning bell rings.  You describe your topology as being a 'router behind a router', so you have two routers operating within your network.  Now, most routers enable their dhcp service by default.  My concern is that, unless you have disabled the dhcp service on one of the routers, you may have two dhcp servers running in your local network.

 

While confusing, this may not, necessarily, be a problem - as long as there is no overlap between the range of addresses that each server will allocate.  However, if one of the routers allocates an address to your unRAID server, which has already been allocated to another machine by the other router, then you will have big problems!

 

However, I wouldn't expect that these problems would include the 'link beat lost' errors which you find in your log.

 

To prove whether this is a problem caused by the LinkSys/unRAID, try disconnecting the LinkSys from your other router and see whether all your machines can make lan connections via the LinkSys without problems.

 

I've just googled AT&T RG - I see that it is provided by your isp.  I guess that the configuration of this device will be 'locked down' and you will have little, if any, control over it.  I presume that you are using the LinkSys to expand your available lan ports and/or enable Gb lan, because you happened to have it 'lying around'.  If this is the case, then yes - you would be better off using a simple switch.  However, as you suggest, it may be possible to reconfigure the LinkSys so that its router functions and services are all disabled.

Yeah, as far as the router behind a router thing, the Uverse gateway is a 2wire3800, it is connected directly to the ISP with ethernet over coax.  It exists at 192.168.1.254.  Its DHCP server only operates from 192.168.1.64 (the phone/dvr) to 192.168.1.163.  I connected the E3000 to the router and placed it in the DMZplus on the gateway.  I set the E3000 on a different subnet, ie 192.168.2.1, and had it hand out IPs with DHCP from 192.168.2.100 to 149.  This was apparently a problem somehow.  You are absolutely correct that I was using the linksys to enable gigE.  Specifically, I wanted it from the server to the HTPC so I could watch movies and copy files and have my wife be able to read jezebel or gawker and whatever at the same time.  Such foolish goals. 

 

So after I connected the server directly to the uverse gateway, it popped right up.  So i used the dd-wrt tutorial to set the linksys up as just a switch.  No wan, no wlan, just wire switch.  When I connected the server to it, it worked at first, but sometime after I got home, it dropped off again.  I'm at work now, and I didn't really have a chance to read the syslog yet, but it seems that it just drops its address occasionally and then never got it back the last time.  No matter, when I connected it back direectly to the gateway, everything worked again. 

 

I ordered the trendnet 8 port switch from amazon, so today I'll get to try that out.

 

As far as the suggestion to switch to real static IPs:  I tried that every time I had a problem.  I'd set the static IP somewhere outside the DHCP range and reboot the server with the new network.cfg, and ifconfig told me everything was good, and it had the right IP, but it was fibbing as far as its connectivity.  Pings wouldn't work, it really wasn't on the network at all.  I don't know, maybe I did it wrong or there's some setting that says "psst, work with slackware" that I can't find.

 

Interestingly, as far as the HTPC having limited connectivity, after I set the e3000 up as just a switch, it works just fine.  But the server doesn't. 

 

My thinking on how to set the network up in the house keeps changing based on this issue.  At first, I had the windows PC doing all the downloading, renaming and tagging, and then copying to the server.  unfortunately, it's an older laptop with only a 10/100 card, so copying to the server took forever.  So I set up sab and sickbeard on the cache drive, and just used the PC for renaming and tagging.  This is, in itself a pain in the a$$ cause there seems to be no good way to move files from the "no mover" zone of the cache drive to the "mover" zone without going back up to the PC and back.  Last night I modified the mover script to do that before it moves stuff, but I don't know how sustainable that is.  It used to work without such a mod, but that's for another thread I guess.

 

I hope that when I get home today the server is still up on the network, and that i've isolated the problem to the E3000.  Though that does leave me with a $100+ linksys paperweight.  It amazes me that it works with every other machine in the house except the ones I need it to!

 

Thanks all, I'm sure you'll all be on the edges of your seats for the next episode... 

 

... i used the dd-wrt tutorial to set the linksys up as just a switch.  No wan, no wlan, just wire switch.

 

So everything was working in the 192.168.1.0/24 subnet?

 

When I connected the server to it, it worked at first, but sometime after I got home, it dropped off again.

 

You are aware that addresses supplied by dhcp are on a limited time lease, and have to be renewed every so often?  It seems as though it may be this renewal which is failing.

 

I ordered the trendnet 8 port switch from amazon, so today I'll get to try that out.

 

I would hope that it will all work properly using the trendnet in place of your Linksys.

 

As far as the suggestion to switch to real static IPs:  I tried that every time I had a problem.  I'd set the static IP somewhere outside the DHCP range and reboot the server with the new network.cfg, and ifconfig told me everything was good, and it had the right IP, but it was fibbing as far as its connectivity.  Pings wouldn't work, it really wasn't on the network at all.  I don't know, maybe I did it wrong or there's some setting that says "psst, work with slackware" that I can't find.

 

I'm pretty sure that your last suggestion isn't the answer!  However, if nothing worked with a static ip configuration, it does suggest that the configuration was at fault.

 

... there seems to be no good way to move files from the "no mover" zone of the cache drive to the "mover" zone without going back up to the PC and back.

 

It may not be a 'good way', but using cli on a telnet session would eliminate the data going over the network.

 

 

If you're talking about moving from one directory on the cache drive to another directory requiring the network then you are doing it wrong. You should be able to move files (not copy but move) on the cache drive in seconds using another machine with no need to have the data go across the network.

 

But then I have no idea why you need to do any file moving or renaming from your PC. Sickbeard and SAB work together to do the sorting and renaming very well all by themselves.

 

Peter

 

@PeterB:  Everything was working until it wasn't anymore.  Although, I will be honest, I don't exactly know what 192.168.1.0/24 is.  Possibly part of my problem.  I am aware of the lease renewal being the problem, I guess I just don't know how to fix it, since sometimes it works, other times it doesn't!

 

@lionelhutz:  As I said, this setup has evolved quite a bit from when I started.  When I switched from transmission/utorrent to sab, I was already using theRenamer and Media Companion on my windows pc.  I have not yet learned how to set SAB and sickbeard up properly cause the other programs WERE working acceptably.  Though the automation could use some work.  I digress.

 

I'll be, assuming everything works when I get home, tutorializing myself in sick and sab this afternoon though, I realize that's the way to go.

I will be honest, I don't exactly know what 192.168.1.0/24 is.

 

It's just a shorthand way of specifying a subnet address range, eg: 192.168.1.0 mask 255.255.255.0

hey, lionelhutz...

 

While I agree I SHOULD be able to move files from the protected, nonmover area of the cache drive, I just tried it again, from my SAB download directory to the movies share (drag and drop), and it runs up over the network at dismal speed.  What am I doing wrong?  I should mention I am using the dreaded macintosh.

hey, lionelhutz...

 

While I agree I SHOULD be able to move files from the protected, nonmover area of the cache drive, I just tried it again, from my SAB download directory to the movies share (drag and drop), and it runs up over the network at dismal speed.  What am I doing wrong?  I should mention I am using the dreaded macintosh.

 

A move is only going to fast if the source and destination are on the same disk. If the source and destination are on different disks then the data has to cross the network twice. Speed disk to disk transfers by using the mover, command line copy, or midnight commander.

Try moving from the cache disk to a Movies directory (named the same as the movies share) on the cache disk.

 

The movie is moved to the movies share on cache disk, right?

 

Peter

Update.

 

The cache drive issue has to take a back seat.

 

After the linksys as a switch method failed, I mentioned I would try the trendnet 8 port switch to enable gigE speeds.  I had the server quite inelegantly plugged into the 100mbit router, but it was working fine.  I decided to try the switch so I could clean up the office and living room of ethernet cables. 

 

STUPID.

 

Doesn't work with the switch.  Won't work plugged back into the router either.  I am now royally screwed.  Ive rebooted, tried a static IP on the switch and the router, and no go.  The only thing left is to install a nic into the server I guess, but it works for everything else like IPMI, so why would that be the problem. 

 

I am so tired of this issue.  Anything you've got at this point folks.  unRAID voodoo doll?  Spells?  Batman?

 

 

 

any reason why you don't disable one of the nics in the bios ?

 

from your "latest" syslog

 

Mar 28 14:13:46 Tower kernel: 0000:03:00.0: eth0: (PCI Express:2.5GB/s:Width x1) 00:25:90:33:5d:ea

Mar 28 14:13:46 Tower kernel: 0000:03:00.0: eth0: Intel® PRO/1000 Network Connection

Mar 28 14:13:46 Tower kernel: 0000:03:00.0: eth0: MAC: 3, PHY: 8, PBA No: 0101ff-0ff

Mar 28 14:13:46 Tower kernel: e1000e 0000:04:00.0: PCI INT A -> GSI 17 (level, low) -> IRQ 17

Mar 28 14:13:46 Tower kernel: e1000e 0000:04:00.0: setting latency timer to 64

Mar 28 14:13:46 Tower kernel: e1000e 0000:04:00.0: irq 35 for MSI/MSI-X

Mar 28 14:13:46 Tower kernel: e1000e 0000:04:00.0: irq 36 for MSI/MSI-X

Mar 28 14:13:46 Tower kernel: e1000e 0000:04:00.0: irq 37 for MSI/MSI-X

Mar 28 14:13:46 Tower kernel: 0000:04:00.0: eth1: (PCI Express:2.5GB/s:Width x1) 00:25:90:33:5d:eb

Mar 28 14:13:46 Tower kernel: 0000:04:00.0: eth1: Intel® PRO/1000 Network Connection

Mar 28 14:13:46 Tower kernel: 0000:04:00.0: eth1: MAC: 3, PHY: 8, PBA No: 0101ff-0ff

 

2 nics detected

 

he can not find uplink on ETH0

 

Mar 28 14:13:46 Tower ifplugd(eth0)[1622]: ifplugd 0.28 initializing.

Mar 28 14:13:46 Tower ifplugd(eth0)[1622]: Using interface eth0/00:25:90:33:5D:EA with driver <e1000e> (version: 1.0.2-k2)

Mar 28 14:13:46 Tower ifplugd(eth0)[1622]: Using detection mode: SIOCETHTOOL

Mar 28 14:13:46 Tower ifplugd(eth0)[1622]: Initialization complete, link beat not detected.Mar 28 14:13:46 Tower ifplugd(eth0)[1622]: ifplugd 0.28 initializing.

Mar 28 14:13:46 Tower ifplugd(eth0)[1622]: Using interface eth0/00:25:90:33:5D:EA with driver <e1000e> (version: 1.0.2-k2)

Mar 28 14:13:46 Tower ifplugd(eth0)[1622]: Using detection mode: SIOCETHTOOL

Mar 28 14:13:46 Tower ifplugd(eth0)[1622]: Initialization complete, link beat not detected.

3 seconds later the link is up

but after apcupsd initialises it goes down again

 

Mar 28 14:13:49 Tower kernel: e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX

Mar 28 14:13:50 Tower ifplugd(eth0)[1622]: Link beat detected.

Mar 28 14:13:51 Tower ifplugd(eth0)[1622]: Executing '/etc/ifplugd/ifplugd.action eth0 up'.

Mar 28 14:13:51 Tower ifplugd(eth0)[1622]: client: Polling for DHCP server on interface eth0:

Mar 28 14:13:52 Tower logger: /etc/rc.d/rc.inet1:  /sbin/dhcpcd -d -t 30 -h Tower eth0

Mar 28 14:13:52 Tower dhcpcd[3046]: broadcasting DHCP_DISCOVER

Mar 28 14:13:55 Tower kernel: e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX

Mar 28 14:13:56 Tower apcupsd[6661]: apcupsd 3.14.3 (20 January 2008) slackware startup succeeded

Mar 28 14:13:56 Tower apcupsd[6661]: NIS server startup succeeded

Mar 28 14:13:56 Tower init: Re-reading inittab

Mar 28 14:14:22 Tower dhcpcd[3046]: timed out waiting for a valid DHCP server response

Mar 28 14:14:22 Tower ifplugd(eth0)[1622]: client: dhcpcd: MAC address = 00:25:90:33:5d:ea

Mar 28 14:14:22 Tower ntpd[8041]: ntpd [email protected] Wed Jan 14 23:46:25 UTC 2009 (1)

Mar 28 14:14:22 Tower ifplugd(eth0)[1622]: client: Starting NTP daemon:  /usr/sbin/ntpd -g

Mar 28 14:14:22 Tower ntpd[8042]: precision = 1.000 usec

Mar 28 14:14:22 Tower ntpd[8042]: ntp_io: estimated max descriptors: 1024, initial socket boundary: 16

Mar 28 14:14:22 Tower ntpd[8042]: Listening on interface #0 wildcard, 0.0.0.0#123 Disabled

Mar 28 14:14:22 Tower ntpd[8042]: Listening on interface #1 lo, 127.0.0.1#123 Enabled

Mar 28 14:14:22 Tower ntpd[8042]: kernel time sync status 0040

Mar 28 14:14:22 Tower ntpd[8042]: frequency initialized 131.070 PPM from /etc/ntp/drift

Mar 28 14:14:22 Tower ifplugd(eth0)[1622]: Program executed successfully.

Mar 28 14:14:22 Tower ifplugd(eth0)[1622]: Link beat lost.

Mar 28 14:14:24 Tower ntpd_initres[8045]: host name not found: pool.ntp.org

Mar 28 14:14:24 Tower ntpd_initres[8045]: couldn't resolve `pool.ntp.org', giving up on it

Mar 28 14:14:25 Tower kernel: e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX

Mar 28 14:14:26 Tower ifplugd(eth0)[1622]: Link beat detected.

it fails ntpd update but after that it is up again ...

 

do you have an UPS connected to your system ?

if so disconnect that to see if the problem goes away ...  seen weird thing happening with networks ups's 

if this is not the cause ...  try putting off npi and second nic ... just to see if it is better ...

if it is better then try to enable ipmi on a dedicated nic not on a shared nic (no clue on supermicro if this is even used but on hp servers i have seen issues with that)

 

for me it seems your network gets interrupted by something...  (duplicate ip/ duplicate mac/ unraid maybe getting confused by the 2 nics ... )

so i would eliminate all extra's and then see how it works

 

only my 2 cents

 

To add to the above. You should also be able to disable the one it's detecting as eth0 so that you can try using the second port just in case the first port has a hardware issue.

 

Peter

I do have a UPS connected.  I didn't at one point and had a bad shutdown, so in it goes.  The problems didn't start until well after that though.  There doesn't seem to be an option in the BIOS to disable either of the NICs.  I did disable the second one by jumpering, and when I booted back up, it worked for I dunno, 12 hours until this morning in the middle of a movie it failed again.  I'd set it up with a static IP too.  Still failed.  Discovered a foolproof way of getting it to restart though, which is to unplug it after it's shut down.  Seems there might be some problem since all the ports stay powered while the thing is shut off.  THere's a setting or a hardware thing or something that gets fixed by unplugging it.

 

I have no desire, however, to shutdown, unplug and restart my server every twelve hours.  Going to try the second port today I guess, I just jumper out the LAN1 port and connect to LAN2?

 

Also, does anyone with this board know how to disable the ports in the BIOS instead?

 

Thanks!

I should also mention, I set this thing up with a static ip, as I said, and since then, it hasn't been able to connect to the internet at all.  It's on the internal network, but can't ping named or numbered addresses outside of the house.

 

Maybe I have the wrong gateway or something, but I just set it to the router...

This sort of thing happened to me with this board when I enabled active power state management. I would lose the network after a period of time and could only get it back after rebooting. Check to see it that is disabled. It is located in the advance tab under advanced chipset control.

 

On a side note you are correct, there is no bios option to disable the extra network controller. It must be disabled on the board.