Hi,
1. I'm trying to start 1000 jails for demonstration purposes (a
presentation), but I can reliably panic the kernel when somewhere
between 400 and 600 jails are started. The panic is not always of the
same type nor at the same place in code, and the number of created jails
also varies. I'm using ZFS for the jails directory and have about 10
nullfs-mounted UFS directories per jail. I'm not sure this is ZFS's
fault but will test with UFS-only soon. I don't have coredumps (because
of 2.) but I have some photos here:
http://ivoras.sharanet.org/stuff/panic/ - the first 3 images are from
one panic, the next two from the next one, the last for the 2nd problem.2. I can't set a kernel coredump directory. The setup is not unusual (no
gmirror, etc.) - simply partitions handled by GPART. The error is in
ioctl(DIOCSKERNELDUMP): Device not configured. The only thing that comes
to my mind is that the swap device is 2 GB while the machine has 4 GB
RAM but AFAIK minidumps should work? If I can get coredumps to work I
could provide more information on the first problem.3. Turning Scroll-lock on (for the purpose of taking the last photo)
before the login prompt (getty) is started makes the console wedge after
the Scroll-lock is released. No input or output is processed though some
keys work (Scroll-lock is one of them).For the first problem, I can share the script that generates the jails,
if anyone wants it.All this is on 8-CURRENT from today. Kernel and world are in sync.
Hi,
In which case, given how memory-hungry ZFS is, it looks like your
problem is with exhausting some memory resource.--
Bob Bishop
rb@gid.co.uk_______________________________________________
freebsd-current@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"
It did occur to me but if it is, why didn't it fail in an obvious way
(e.g. kmem_map panics with ZFS), and how to tune it so if works? This is
AMD64 with 0.5 GB - 1 GB memory in the Free bin when it panics.
I'd suggest increasing kernel mem but I don't know if/how that's
tunable on amd64.--
Bob Bishop
rb@gid.co.uk_______________________________________________
freebsd-current@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"
What's the partition type of the partition you're trying to dump
to?--
Marcel Moolenaar
xcllnt@mac.com_______________________________________________
freebsd-current@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"
It's ad8s2b from our earlier conversation.
2. Name: ad8s2b
Mediasize: 2147483648 (2.0G)
Sectorsize: 512
Mode: r1w1e0
rawtype: 7
length: 2147483648
offset: 19327352832
type: freebsd-ufs
index: 2
end: 41943039
start: 37748736I see it's labeled as freebsd-ufs. AFAIK I created this file system with
sysinstall.Do I need to change the type? Any special instructions?
Yes. Change the type to freebsd-swap if you're going to be
dumping kernel core files on it. The anti-footshooting
checks in gpart prevent you from dumping to UFS partitions:% sudo gpart modify -i 2 -t freebsd-swap ad8s2
FYI,
--
Marcel Moolenaar
xcllnt@mac.com_______________________________________________
freebsd-current@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"
Yes.
change the type as follows:
sudo gpart modify -t freebsd-swap -i 2 ad8s2bdumpon should work then.
--
Paul
_______________________________________________
freebsd-current@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"
ZFS doesn't influence it - the same problem happens without ZFS on the
system (though I can create a few hundred more jails without ZFS before
the kernel panics...). I've uploaded more photos.
| Ingo Molnar | [patch 12/13] syslets: x86: optimized copy_uatom() |
| Greg Kroah-Hartman | [PATCH 017/196] aoechr: Convert from class_device to device |
| Yinghai Lu | Re: 2.6.26, PAT and AMD family 6 |
| Jan Engelhardt | intel iommu (Re: -mm merge plans for 2.6.23) |
git: | |
| Gerrit Renker | [PATCH 27/37] dccp: Integration of dynamic feature activation - part 2 (server side) |
| David Miller | [GIT]: Networking |
| David Miller | Re: [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| Natalie Protasevich | [BUG] New Kernel Bugs |
