On 09/27/2012 11:00 AM, Rudy wrote:
On 09/25/2012 01:37 AM, Hooman Fazaeli wrote:
dev.em.1.link_irq: 6379725883
dev.em.2.link_irq: 6379294926
Based on the strangely high value of dev.em.1.link_irq (which means too
many link
status changes: down -> up -> down -> ....), I guess the problem is the
same as
discussed in these threads:
http://lists.freebsd.org/pipermail/freebsd-net/2011-November/030424.html
http://lists.freebsd.org/pipermail/freebsd-net/2012-March/031648.html
To confirm, you may run this test:
1. Start a ping flood: ping -f <other-machine-ip>
2. Let it run for a few seconds.
3. Disconnect the cable.
4. After a while, you should see "no buffer space" error.
5. Stop ping flood.
6. Re-connect the cable and wait 10 seconds.
7. Start a normal ping. Error messages should show up again.
To fix, upgrade to the latest e1000 driver from HEAD.
The very high link_irq may be due to a loose connection.
Replace the patch cord and see if it helps.
Thanks for the tips. I will test next time I am at the data center. For
now, I rebooted after doubled the default nmbclusters and quadrupled the
hw.em.rxd values in loader.conf.
# loader.conf
kern.ipc.nmbclusters=524288
hw.igb.rxd=4096
hw.igb.txd=4096
hw.em.rxd=4096
hw.em.txd=4096
Rebooting and/or the settings change seems to have stopped the errors.
Here is a pretty little graph showing error rate on em1 for the past 3
days.
http://www.monkeybrains.net/images/ErrorRate-em1.png
Rudy
_______________________________________________
freebsd-net@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-net
To unsubscribe, send any mail to "freebsd-net-unsubscr...@freebsd.org"
_______________________________________________
freebsd-net@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-net
To unsubscribe, send any mail to "freebsd-net-unsubscr...@freebsd.org"
