[Bugme-new] [Bug 10890] New: Heavy iSCSI traffic triggers memory allocation complaints and system lockup
bugme-daemon at bugzilla.kernel.org
bugme-daemon at bugzilla.kernel.org
Tue Jun 10 02:56:46 PDT 2008
http://bugzilla.kernel.org/show_bug.cgi?id=10890
Summary: Heavy iSCSI traffic triggers memory allocation
complaints and system lockup
Product: Other
Version: 2.5
KernelVersion: 2.6.25.6
Platform: All
OS/Version: Linux
Tree: Mainline
Status: NEW
Severity: high
Priority: P1
Component: Other
AssignedTo: other_other at kernel-bugs.osdl.org
ReportedBy: bart.vanassche at gmail.com
Latest working kernel version: (not known)
Earliest failing kernel version: fails at least with 2.6.24.3-ubuntu and
2.6.25.6
Distribution: Ubuntu Linux 8.04 server
Hardware Environment: x86_64
Software Environment:
Problem Description:
System starts behaving strange after about an hour of heavy iSCSI traffic: the
console locks up (keys typed in are not echoed) and error messages are logged
about problems with memory allocation.
Steps to reproduce:
This is how my setup looks like:
- One server is running OpenSolaris build 90 and there have been 64 iSCSI
targets configured on this system, where each target is 10 GB in size.
- On the second server I have installed Ubuntu 8.04 server 64-bit, and I
replaced the Ubuntu kernel by the 2.6.25.6 kernel.
- Both systems are connected via a 4x SDR InfiniBand network (10 Gbit/s).
- Open-iscsi was configured on the Linux server such that it connects to all 64
iSCSI targets and such that I/O to all 64 targets was active simultaneously. I
used the following command (see below for the scripts):
./iscsi-logout-and-login && ./iscsi-stress-test &
# cat iscsi-logout
#!/bin/bash
target_ip="${1:-192.168.67.10}"
iscsiadm -m session \
| while read proto num ip iqn
do
iscsiadm -m node -p ${ip} -T $iqn --logout
done
# cat iscsi-logout-and-login
#!/bin/bash
target_ip=192.168.67.10
./iscsi-logout "${target_ip}"
i=0
iscsiadm -m discovery -t st -p ${target_ip} \
| grep '\.67\.' \
| cut -f2 -d' ' \
| sort -k 1.60 \
| while read iqn
do
if [ "$1" != "" -a "$i" -ge "${1:-0}" ]; then
break
fi
iscsiadm -m node -p ${target_ip} -T ${iqn} --login
iscsiadm --mode node -p ${target_ip} -T ${iqn} \
--op update -n "node.session.timeo.replacement_timeout" -v 10
iscsiadm --mode node -p ${target_ip} -T ${iqn} \
--op update -n "node.conn[0].timeo.noop_out_interval" -v 0
iscsiadm --mode node -p ${target_ip} -T ${iqn} \
--op update -n "node.conn[0].timeo.noop_out_timeout" -v 0
i=$((i+1))
done
# cat iscsi-stress-test
#!/bin/bash
while true
do
date
for d in /dev/disk/by-path/ip-*:3260-iscsi-iqn.1986-03.com.sun*
do
dd if=/dev/zero of=$d bs=100K &
done
wait
done
--
Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.
More information about the Bugme-new
mailing list