pfSense

Well well well :)
https://forum.pfsense.org/index.php?topic=107471.msg602590#msg602590

See last message: https://forum.pfsense.org/index.php?topic=107471.msg603180#msg603180 it helped me since then.

This has been happening to me, the OUTBOUND traffic on the LAN interface continues to work but the driver or something is stuck and no data is ever read off the NIC. This was especially bad becuase it would always be the CARP master that died, and the CARP advertisments had no problem going out to the secondary thereby blocking CARP failover. I couldn't figure out how to troubleshoot it and just went back to 2.2 but it's happened on another 2.3 instance today. vmxnet3 does the same but it goes into 100% interrupt wait. I verified this on ESX where no data was ever handed off to the driver on both host and VM counters. I gave up primarily because I can't figure out how to duplicate it in a timely fashion. I'd like to help fix this if there is anything that would be helpful... ie debug on the VM etc - if there is some way to duplicate it now?

my test setups are past 4 hours run time now, in a scenario where they never exceeded 4 hours before the fix.

One user who could easily replicate in minutes every time has confirmed it's no longer happening for him either.

needs more run time, but initial results look promising.

all still running fine after around 11 hours. That's more than 7 hours longer than any affected kernel has lasted in my current stress test.

Chris Buechler wrote:

This has lasted over 3 days now in multiple scenarios that wouldn't last 4 hours on affected kernels. Multiple outside reports from people who could more frequently replicate than most that it's also fixed for them.

Is there a schedule for this patch to be released or do I have to switch to the developer snapshot cycle?

Brent Kerlin wrote:

Chris Buechler wrote:

This has lasted over 3 days now in multiple scenarios that wouldn't last 4 hours on affected kernels. Multiple outside reports from people who could more frequently replicate than most that it's also fixed for them.

Is there a schedule for this patch to be released or do I have to switch to the developer snapshot cycle?

Its in the dev snapshots & it'll be in 2.3.1, which will be released soon (somewhere the next couple of days if nothing goes wrong)

Thanks guys!

For the official boxes (ADI image) there is no 2.3.1 development branch that we could upgrade to correct? So until 2.3.1 is released either live with the problem or disable the additional cores?

Regards,
JP

This just happened to me again on 2.3.1 - I've left the system in the locked up state in case you want (or want me) to check anything on it.

I am seeing issued similar or the same as posted in this thread with 2.3.2 (2.3.1 was a typo), SMP, and IPsec. CPU randomly hits 100%. When I disable SMP the problem goes away. How do we re-open this case?

Version: 2.3.2-RELEASE (amd64)
built on Tue Jul 19 12:44:43 CDT 2016
FreeBSD 10.3-RELEASE-p5

CPU Type: Intel(R) Atom(TM) CPU D2550 @ 1.86GHz
Load average: 2.60, 2.25, 2.03
CPU usage: 86%
Memory usage: 6% of 4050 MiB

Not enough info there. Try again on 2.3.2, post on the forum, try to narrow it down. Unlikely to be the same bug, but there is a small chance it may be related.