Re: High load - lost network
Jeremy Chadwick wrote: I'm not really sure anyone will know how to fix this. Sometimes a BIOS upgrade can fix such things, other times motherboard replacements are in order. I'll check the BIOS and see if I could do something with it. If I enable POLLING, could that fix the problem? -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Jeremy Chadwick wrote: Simply put: I don't know. Based on the polling(4) man page, it might improve things for you, but your ATA high interrupt rate problem will still exist even if you use polling(4). Okay. Thanks for taking time helping me :) -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
On Tue, Nov 11, 2008 at 09:00:18AM +0100, Christer Solskogen wrote: Jeremy Chadwick wrote: I'm not really sure anyone will know how to fix this. Sometimes a BIOS upgrade can fix such things, other times motherboard replacements are in order. I'll check the BIOS and see if I could do something with it. If I enable POLLING, could that fix the problem? Simply put: I don't know. Based on the polling(4) man page, it might improve things for you, but your ATA high interrupt rate problem will still exist even if you use polling(4). -- | Jeremy Chadwickjdc at parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP: 4BD6C0CB | ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Uit een eerder bericht (10-11-2008 13:45): Do anyone have a tips for how to workaround this or is the server just junk? Got this problem a while ago as well. I found out with me it was hub related. If I was downloading, my hub was displaying 100Mbit connection (I wish it was :-) Then I sometimes get this nic status of my BSD system as well. I bought a 3Com hub and the problem is gone. regards, Jos Chrispijn ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Christer Solskogen wrote: I have a server with a em interface. Whenever the server has a high load (compiling world for instance) the network connectivity is lost. The solution was to disable the onboard network card, and insert a pci-x card instead. -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Jos Chrispijn([EMAIL PROTECTED])@2008.11.11 09:37:47 +0100: Uit een eerder bericht (10-11-2008 13:45): Do anyone have a tips for how to workaround this or is the server just junk? Got this problem a while ago as well. I found out with me it was hub related. If I was downloading, my hub was displaying 100Mbit connection (I wish it was :-) Then I sometimes get this nic status of my BSD system as well. I bought a 3Com hub and the problem is gone. Must have been one of those cisco that they bought from a different company, then relabeled (forget which). Man was that a POS. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
High load - lost network
Christer Solskogen writes: I have a server with a em interface. Whenever the server has a high load (compiling world for instance) the network connectivity is lost. dmesg tells me this: em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN Have you double-checked the hardware? (Includes the cable connection.) The em driver has a very good reputation and others - includong myself - use it under siilar load profiles with no problems. Robert Huff ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Robert Huff wrote: Have you double-checked the hardware? (Includes the cable connection.) The em driver has a very good reputation and others - includong myself - use it under siilar load profiles with no problems. Yeah, I have. The error is replicated just by giving the machine something to do. It might also have something to do with the auto neogation and the switch. the wierd thing is that the only message in the log is: Nov 10 13:29:58 caius kernel: em0: link state changed to DOWN Nov 10 13:29:59 caius kernel: em0: link state changed to UP Nov 10 13:30:08 caius kernel: em0: link state changed to DOWN Nov 10 13:30:10 caius kernel: em0: link state changed to UP Nothing else :/ -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
On Monday 10 November 2008 15:18:40 Christer Solskogen wrote: Robert Huff wrote: Have you double-checked the hardware? (Includes the cable connection.) The em driver has a very good reputation and others - includong myself - use it under siilar load profiles with no problems. Yeah, I have. The error is replicated just by giving the machine something to do. It might also have something to do with the auto neogation and the switch. the wierd thing is that the only message in the log is: Nov 10 13:29:58 caius kernel: em0: link state changed to DOWN Nov 10 13:29:59 caius kernel: em0: link state changed to UP Nov 10 13:30:08 caius kernel: em0: link state changed to DOWN Nov 10 13:30:10 caius kernel: em0: link state changed to UP Any shared interrupts? What does vmstat -i show under the load you describe? -- Mel Problem with today's modular software: they start with the modules and never get to the software part. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
At 06:45 AM 11/10/2008, Christer Solskogen wrote: I have a server with a em interface. Whenever the server has a high load (compiling world for instance) the network connectivity is lost. dmesg tells me this: em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP Do anyone have a tips for how to workaround this or is the server just junk? The server is running FreeBSD 7.0 (i386) with latest patches from freebsd-update. -- chs Check your switch and the wiring. I use em drivers and hardware without issues. -Derek -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Mel wrote: Any shared interrupts? What does vmstat -i show under the load you describe? irq1: atkbd0 531 0 irq6: fdc010 0 irq14: ata0 95 0 irq18: em0 atapci1 1198845 4 cpu0: timer559484003 1999 cpu1: timer559483898 1999 Total 1120167382 4004 -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
On Monday 10 November 2008 15:44:59 Christer Solskogen wrote: irq18: em0 atapci1 1198845 4 Since disk and network card share the same IRQ, does the problem occur if you generate a similar load, without using disk IO? Like, put a big file on a memory disk and bzip it. Could be a red herring, but good to rule out. -- Mel Problem with today's modular software: they start with the modules and never get to the software part. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
On Mon, Nov 10, 2008 at 01:45:44PM +0100, Christer Solskogen wrote: I have a server with a em interface. Whenever the server has a high load (compiling world for instance) the network connectivity is lost. dmesg tells me this: em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP Do anyone have a tips for how to workaround this or is the server just junk? Can you provide dmesg | grep em0 output? I'd like to see what version of NIC this is. -- | Jeremy Chadwickjdc at parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP: 4BD6C0CB | ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Christer Solskogen([EMAIL PROTECTED])@2008.11.10 13:45:44 +0100: I have a server with a em interface. Whenever the server has a high load (compiling world for instance) the network connectivity is lost. I have a problem where the whole machine becomes unresponsive on sustained disk IO every few seconds. 7.1-BETA2. No shared irqs. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Jeremy Chadwick wrote: Can you provide dmesg | grep em0 output? I'd like to see what version of NIC this is. No problem. em0: Intel(R) PRO/1000 Network Connection Version - 6.7.3 port 0xbc00-0xbc1f mem 0xfc5e-0xfc5f irq 18 at device 1.0 on pci1 em0: Ethernet address: 00:02:b3:ea:28:f0 em0: [FILTER] -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
High load - lost network
I have a server with a em interface. Whenever the server has a high load (compiling world for instance) the network connectivity is lost. dmesg tells me this: em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN em0: link state changed to UP Do anyone have a tips for how to workaround this or is the server just junk? The server is running FreeBSD 7.0 (i386) with latest patches from freebsd-update. -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
em0: link state changed to DOWN em0: link state changed to UP em0: link state changed to DOWN Have you double-checked the hardware? (Includes the cable connection.) The em driver has a very good reputation and others - includong myself - use it under siilar load profiles with no and including myself. intel card+em driver are perfect. this em0: link state changed to UP em0: link state changed to DOWN suggest switch and/or cable problem. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
i don't think it's interrupt sharing problems. anyway - turn on MSI interrupts if your hardware can. check if producing high load on atapci1 make problem worse or not. On Mon, 10 Nov 2008, Christer Solskogen wrote: Mel wrote: Any shared interrupts? What does vmstat -i show under the load you describe? irq1: atkbd0 531 0 irq6: fdc010 0 irq14: ata0 95 0 irq18: em0 atapci1 1198845 4 cpu0: timer559484003 1999 cpu1: timer559483898 1999 Total 1120167382 4004 -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
On Mon, Nov 10, 2008 at 05:52:59PM +0100, Christer Solskogen wrote: Jeremy Chadwick wrote: Can you provide dmesg | grep em0 output? I'd like to see what version of NIC this is. No problem. em0: Intel(R) PRO/1000 Network Connection Version - 6.7.3 port 0xbc00-0xbc1f mem 0xfc5e-0xfc5f irq 18 at device 1.0 on pci1 em0: Ethernet address: 00:02:b3:ea:28:f0 em0: [FILTER] Well shoot, that didn't tell me what I want. pciconv -lv, and look for the em0 entry? (I need all the lines shown associated with it) Thanks. -- | Jeremy Chadwickjdc at parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP: 4BD6C0CB | ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Jeremy Chadwick wrote: Well shoot, that didn't tell me what I want. pciconv -lv, and look for the em0 entry? (I need all the lines shown associated with it) Here you go: [EMAIL PROTECTED]:1:1:0: class=0x02 card=0x342f8086 chip=0x10758086 rev=0x00 hdr=0x00 vendor = 'Intel Corporation' device = '82547EI Gigabit Ethernet Controller' class = network subclass = ethernet -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
Wojciech Puchar wrote: i don't think it's interrupt sharing problems. anyway - turn on MSI interrupts if your hardware can. I dont seem so (unless i interpret the output of pciconf wrong: [EMAIL PROTECTED] ~]# pciconf -lc [EMAIL PROTECTED]:0:0:0: class=0x06 card=0x342f8086 chip=0x25788086 rev=0x02 hdr=0x00 cap 09[e4] = vendor (length 6) Intel cap 3 version 1 [EMAIL PROTECTED]:0:3:0: class=0x060400 card=0x chip=0x257b8086 rev=0x02 hdr=0x01 [EMAIL PROTECTED]:0:28:0: class=0x060400 card=0x chip=0x25ae8086 rev=0x02 hdr=0x01 cap 07[50] = PCI-X 64-bit bridge supports [EMAIL PROTECTED]:0:29:0: class=0x0c0300 card=0x342f8086 chip=0x25a98086 rev=0x02 hdr=0x00 [EMAIL PROTECTED]:0:29:1: class=0x0c0300 card=0x342f8086 chip=0x25aa8086 rev=0x02 hdr=0x00 [EMAIL PROTECTED]:0:29:4: class=0x088000 card=0x342f8086 chip=0x25ab8086 rev=0x02 hdr=0x00 [EMAIL PROTECTED]:0:29:5:class=0x080020 card=0x342f8086 chip=0x25ac8086 rev=0x02 hdr=0x00 cap 07[50] = PCI-X 64-bit supports 512 burst read, 1 split transaction [EMAIL PROTECTED]:0:29:7: class=0x0c0320 card=0x342f8086 chip=0x25ad8086 rev=0x02 hdr=0x00 cap 01[50] = powerspec 2 supports D0 D3 current D0 cap 0a[58] = EHCI Debug Port at offset 0x80 in map 0x14 [EMAIL PROTECTED]:0:30:0: class=0x060400 card=0x chip=0x244e8086 rev=0x0a hdr=0x01 [EMAIL PROTECTED]:0:31:0: class=0x060100 card=0x chip=0x25a18086 rev=0x02 hdr=0x00 [EMAIL PROTECTED]:0:31:1:class=0x01018a card=0x342f8086 chip=0x25a28086 rev=0x02 hdr=0x00 [EMAIL PROTECTED]:0:31:2:class=0x01048f card=0x34308086 chip=0x25b08086 rev=0x02 hdr=0x00 [EMAIL PROTECTED]:0:31:3: class=0x0c0500 card=0x342f8086 chip=0x25a48086 rev=0x02 hdr=0x00 [EMAIL PROTECTED]:1:1:0: class=0x02 card=0x342f8086 chip=0x10758086 rev=0x00 hdr=0x00 cap 01[dc] = powerspec 2 supports D0 D3 current D0 [EMAIL PROTECTED]:3:0:0: class=0x03 card=0x342f8086 chip=0x47521002 rev=0x27 hdr=0x00 cap 01[5c] = powerspec 2 supports D0 D1 D2 D3 current D0 [EMAIL PROTECTED]:3:1:0:class=0x02 card=0x342f8086 chip=0x12298086 rev=0x10 hdr=0x00 cap 01[dc] = powerspec 2 supports D0 D1 D2 D3 current D0 check if producing high load on atapci1 make problem worse or not. During dd if=/dev/random of=bigfile bs=500k count=5 (about 200 seconds) made the network unresponsive again. Same symptom as I got when running make buildworld. -- chs ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
On Tue, Nov 11, 2008 at 08:13:23AM +0100, Christer Solskogen wrote: Jeremy Chadwick wrote: Well shoot, that didn't tell me what I want. pciconv -lv, and look for the em0 entry? (I need all the lines shown associated with it) Here you go: [EMAIL PROTECTED]:1:1:0: class=0x02 card=0x342f8086 chip=0x10758086 rev=0x00 hdr=0x00 vendor = 'Intel Corporation' device = '82547EI Gigabit Ethernet Controller' class = network subclass = ethernet Thanks. There's a specific model that has a watchdog timeout problem, which is known to cause network drops. The root cause was an improperly set bit inside of the NIC's EEPROM (and can be fixed using a DOS utility from Intel). Your NIC isn't that particular model, so you're safe. Your issue appears to be with the ATA controller on your machine having a very high interrupt rate, and since the NIC's IRQ is shared with that, any heavy interrupt activity causes the opposing device to malfunction. I'm not really sure anyone will know how to fix this. Sometimes a BIOS upgrade can fix such things, other times motherboard replacements are in order. -- | Jeremy Chadwickjdc at parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP: 4BD6C0CB | ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: High load - lost network
On Tuesday 11 November 2008 08:25:10 Christer Solskogen wrote: During dd if=/dev/random of=bigfile bs=500k count=5 (about 200 seconds) made the network unresponsive again. Same symptom as I got when running make buildworld. And same command with of=/dev/null? -- Mel Problem with today's modular software: they start with the modules and never get to the software part. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
