[Bug 309945] [NEW] [jaunty, intrepid] Hard system freeze using open-iscsi

[Christian Roessner]

Public bug reported:

Binary package hint: open-iscsi

Hi,

I installed iscsitarget on my host, and in a kvm guest, I tried to use
my configured target. First my setup:

Host:
Intrepid, latest proposed linux image (with fixes for iscsitarget)

cat /etc/ietd.conf | grep -ve ^.*# -ve ^$
Target iqn.2008-12.de.roessner-net:storage.disk0.testing
	Lun 0 Path=/dev/vg01/lv_testing,Type=fileio
	Alias Test
	HeaderDigest		CRC32C,None
	DataDigest		CRC32C,None

The host is still running fine, with following dmesg-output, after the
guest has been died:

...
[ 2220.173286] vcpu not ready for apic_round_robin
[ 2220.670873] vcpu not ready for apic_round_robin
[ 2220.673095] vcpu not ready for apic_round_robin
[ 2221.171379] vcpu not ready for apic_round_robin
[ 2221.173452] vcpu not ready for apic_round_robin
[ 2278.588125] vcpu not ready for apic_round_robin
...

Does not wonder me, because the guest is dead :-)

The guest:

Jaunty (Former Intrepid with same results):

Here the commands, I set in the guest, directly copied over from the
dead console:

ubuntu /etc/rcS.d # mv S25open-iscsi S45open-iscsi
ubuntu /etc/rcS.d # vi S45open-iscsi 
ubuntu /etc/rcS.d # cd
ubuntu ~ # cd /etc/iscsi/
ubuntu /etc/iscsi # vi iscsid.conf 
ubuntu /etc/iscsi # iscsiadm -m discovery -tst -p 10.253.1.254 -d3
iscsiadm: Max file limits 1024 1024

iscsiadm: starting sendtargets discovery, address 10.253.1.254:3260, 
iscsiadm: connecting to 10.253.1.254:3260
iscsiadm: connected local port 53014 to 10.253.1.254:3260
iscsiadm: connected to discovery address 10.253.1.254
iscsiadm: login response status 0000
iscsiadm: discovery process to 10.253.1.254:3260 exiting
iscsiadm: Could not open /etc/iscsi/send_targets/10.253.1.254,3260 err 2

10.253.1.254:3260,1 iqn.2008-12.de.roessner-net:storage.disk0.testing
ubuntu /etc/iscsi # iscsiadm -m node -T iqn.2008-12.de.roessner-net:storage.disk0.testing -p 10.253.1.254 -l
Logging in to [iface: default, target: iqn.2008-12.de.roessner-net:storage.disk0.testing, portal: 10.253.1.254,3260]
Login to [iface: default, target: iqn.2008-12.de.roessner-net:storage.disk0.testing, portal: 10.253.1.254,3260]: successful
ubuntu /etc/iscsi # 

Crash

I only changed two things in the config file. I set manual to automatic
mount and enabled CRC32 stuff for header and data.

I first had tested this setup on my development server with 4 guests
using one target with ocfs2. On the intrepid nodes, I could mount the
partition and start bonnie++. But after a while, all nodes died. So this
time I tried to set up a little testing scenario here at my home
desktop, to fill this bug and as you can see, it also crashes.

If you can speak german, you may find more detailed information here in
my blog:

http://www.roessner-net.com/?p=243

iSCSI would be a good solution for not using NFS.

** Affects: open-iscsi (Ubuntu)
     Importance: Undecided
         Status: New

-- 
[jaunty, intrepid] Hard system freeze using open-iscsi
https://bugs.launchpad.net/bugs/309945
You received this bug notification because you are a member of Ubuntu
Server Team, which is subscribed to open-iscsi in ubuntu.