Bug 216814 - BRAS at FreeBSD 11.0p7 reboots/crash some days (mpd5 vmware esxi 6.0.0)
Summary: BRAS at FreeBSD 11.0p7 reboots/crash some days (mpd5 vmware esxi 6.0.0)
Status: Closed Not A Bug
Alias: None
Product: Base System
Classification: Unclassified
Component: kern (show other bugs)
Version: 11.0-RELEASE
Hardware: amd64 Any
: --- Affects Only Me
Assignee: freebsd-virtualization (Nobody)
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2017-02-05 09:08 UTC by Konstantin V Bekreyev
Modified: 2017-03-20 06:13 UTC (History)
2 users (show)

See Also:


Attachments
/var/log/messages (16.21 KB, text/plain)
2017-02-05 09:08 UTC, Konstantin V Bekreyev
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Konstantin V Bekreyev 2017-02-05 09:08:23 UTC
Created attachment 179629 [details]
/var/log/messages

I using FreeBSD 11 as BRAS with mpd5-5.8 on VMWare ESXi 6.0.0 (3620759) and system crash and reboots abouts every three days. What can be wrong?

$ freebsd-version -uk
11.0-RELEASE-p2
11.0-RELEASE-p7

Feb  5 02:29:36 BRAS3 kernel: current process           = 12 (irq275: vmx2)
Feb  5 02:29:36 BRAS3 kernel: trap number               = 12
Feb  5 02:29:36 BRAS3 kernel: panic: page fault
Feb  5 02:29:36 BRAS3 kernel: cpuid = 2

vmx2 used for PPPoE customers connected via mpd5

Server: HP Proliant DL380 G7 with Intel Xeon E5630 and Broadcom NC382i GE

In sysctl I change only this: kern.ipc.maxsockbuf=23175936

Attaching /var/log/messages about this reboot.

Another virtual instanced of FreeBSD on this VMware working very well and not reboots.
Comment 1 Shane 2017-02-07 00:06:25 UTC
What version is the stable system? UFS or ZFS? Do both have 4G? What varies with the second system - software/settings?

The panic relates to a page fault, pointing to out of memory.

In general 4G for a system using ZFS is low. Keep an eye on ZFS ARC level, maybe set vfs.zfs.arc_max to keep some ram free for your software.

On my desktop I have seen excess "wired" allocations lead to problems. I get a wired reading using `top -d1 | grep Wired | awk '{print $6}'` Any chance you see over 3G wired?
Comment 2 Andrey V. Elsukov freebsd_committer freebsd_triage 2017-02-07 08:34:57 UTC
(In reply to Shane from comment #1)
> What version is the stable system? UFS or ZFS? Do both have 4G? What varies
> with the second system - software/settings?
> 
> The panic relates to a page fault, pointing to out of memory.

The panic looks unrelated to virtualization. We have a lot of such reports from netgraph users.
Comment 3 Konstantin V Bekreyev 2017-02-07 09:51:53 UTC
(In reply to Shane from comment #1)

yes, these systems have ZFS and above 4G of memory (4G BRAS, 20G another)

I will try to add more then 4G of memory.

# sysctl vfs.zfs.arc_max
vfs.zfs.arc_max: 3046846464

# top -d1 | grep Wired | awk '{print $6}'
1168M
Comment 4 Konstantin V Bekreyev 2017-03-20 06:13:33 UTC
Yes, when I added more memory than 4Gb - uptime looks well.