Posted August 21, 201113 yr I have a network card Intel ® PRO/1000 gigabit and the network performance seems very low. How can I check it precisely?
August 21, 201113 yr Author I specify that the server unRAID and PC are connected to a Gigabit switch.
August 21, 201113 yr What speed are you getting? How big is the test file you are using? Is the transfer over SMB? Do you have another PC on the network..what speed do you get transferring between those two machines (take the unRaid out if the equation for the testing...)
August 21, 201113 yr Author Syslog attached NIC info (from ethtool) Settings for eth0: Supported ports: [ TP ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Supports auto-negotiation: Yes Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Advertised pause frame use: No Advertised auto-negotiation: Yes Speed: 1000Mb/s Duplex: Full Port: Twisted Pair PHYAD: 0 Transceiver: internal Auto-negotiation: on MDI-X: Unknown Supports Wake-on: umbg Wake-on: g Current message level: 0x00000007 (7) Link detected: yes NIC driver info (from ethtool -i) driver: e1000 version: 7.3.21-k8-NAPI firmware-version: N/A bus-info: 0000:08:02.0 Ethernet config info (from ifconfig) eth0 Link encap:Ethernet HWaddr 00:1b:21:a2:34:f6 inet addr:192.168.1.17 Bcast:192.168.1.255 Mask:255.255.255.0 UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:316374 errors:0 dropped:7 overruns:0 frame:0 TX packets:615789 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:28004343 (26.7 MiB) TX bytes:2432000290 (2.2 GiB) syslog-2011-08-21.txt
August 21, 201113 yr Author What speed are you getting? How big is the test file you are using? 5 Go Is the transfer over SMB? Yes Do you have another PC on the network..what speed do you get transferring between those two machines (take the unRaid out if the equation for the testing...) 4 Mb/s but the other pc has a 10/100 card.
August 21, 201113 yr See all the link up/down messages? Your nic is definitely not happy. There was also an unhandled interrupt or something causing the nobody cared trace at 13:19. That's either your nic or something that shares IRQ 17 with it. Can we see the output from: cat /proc/interrupts This doesn't look like a physical link problem but a simple test to get out of the way would be to try another Ethernet cable and another port on the switch/router it's plugged into. What motherboard is this?
August 22, 201113 yr Author See all the link up/down messages? Your nic is definitely not happy. There was also an unhandled interrupt or something causing the nobody cared trace at 13:19. That's either your nic or something that shares IRQ 17 with it. Can we see the output from: cat /proc/interrupts This doesn't look like a physical link problem but a simple test to get out of the way would be to try another Ethernet cable and another port on the switch/router it's plugged into. What motherboard is this? Motherboard : Asus P8H67-V As integrated network adapter is not recognized by unRAID I installed an Intel Pro 1000 Thank You for your help
August 22, 201113 yr Author I rebooted the server with the option irqpool as shown in syslog and I went to over 35 Mb/s
August 22, 201113 yr Author But after some time I again the occurrence of these errors in syslog : Aug 22 13:42:55 HCSRV1 kernel: irq 17: nobody cared (try booting with the "irqpoll" option) Aug 22 13:42:55 HCSRV1 kernel: Pid: 0, comm: swapper Not tainted 2.6.37.6-unRAID #4 Aug 22 13:42:55 HCSRV1 kernel: Call Trace: Aug 22 13:42:55 HCSRV1 kernel: [<c104e088>] __report_bad_irq+0x2e/0x6f Aug 22 13:42:55 HCSRV1 kernel: [<c104e1be>] note_interrupt+0xf5/0x14b Aug 22 13:42:55 HCSRV1 kernel: [<c104cfb4>] ? handle_IRQ_event+0x1d/0x97 Aug 22 13:42:55 HCSRV1 kernel: [<c104e92e>] handle_fasteoi_irq+0x97/0xb5 Aug 22 13:42:55 HCSRV1 kernel: [<c104e897>] ? handle_fasteoi_irq+0x0/0xb5 Aug 22 13:42:55 HCSRV1 kernel: <IRQ> [<c1003c98>] ? do_IRQ+0x3e/0x97 Aug 22 13:42:55 HCSRV1 kernel: [<c1002ce9>] ? common_interrupt+0x29/0x30 Aug 22 13:42:55 HCSRV1 kernel: [<c11d2926>] ? acpi_idle_enter_bm+0x232/0x266 Aug 22 13:42:55 HCSRV1 kernel: [<c128a23d>] ? cpuidle_idle_call+0x72/0xa9 Aug 22 13:42:55 HCSRV1 kernel: [<c1001a89>] ? cpu_idle+0x3a/0x51 Aug 22 13:42:55 HCSRV1 kernel: [<c130b6f4>] ? rest_init+0x58/0x5a Aug 22 13:42:55 HCSRV1 kernel: [<c1499846>] ? start_kernel+0x283/0x288 Aug 22 13:42:55 HCSRV1 kernel: [<c14990ce>] ? i386_start_kernel+0xce/0xd5 On the console I get the message: Disabling IRQ #17
August 22, 201113 yr I found a mention of this in a kernel bug https://bugzilla.redhat.com/show_bug.cgi?id=457648
August 22, 201113 yr As a work-around, it may require to change bios settings, or potentially adding a line to one of the .conf files First it will require to identify which bios option is causing the irq nobody cared fault... Recent info on the bug https://bugzilla.redhat.com/show_bug.cgi?id=713351 Appears to be a problem with the Asus MB https://bugzilla.kernel.org/show_bug.cgi?id=39122#c3 http://vip.asus.com/forum/view.aspx?board_id=1&model=E35M1-M+PRO&id=20110507053520320&page=1&SLanguage=en-us
August 22, 201113 yr Author I put the bios update and after 30 minutes the error returned :'( I am desperate
August 22, 201113 yr Author From the bug reports...limited success with "acpi=off" In the boot options?
August 22, 201113 yr yes. Don't know much more than that as a quick search reveals many reports of this irq error and the Asus SandyBridge MB http://www.mentby.com/chris-palmer-2/ http://forums.vr-zone.com/hardware-depot/1437078-asus-p8h67-v-intel-gt-1000pro-nic-fedora-14-getting-very-slow-network-speeds.html http://forums.overclockers.co.uk/showthread.php?t=18285181 and more.
August 23, 201113 yr Getting back to the OP, how's the net performance, both reading and writing? If you get a chance, try pci=nomsi instead of the noacpi. It completely depends on how the configuration is settling out but one or the other may be faster.
August 23, 201113 yr Author With acpi=off reading and writing : 25-30 Mb/s With pci=nomsi errors appear
Archived
This topic is now archived and is closed to further replies.