native_thread process takes 100% CPU

[Ian Lochray]

We have got a VA71100 SAN Disk Array. There are a number of native_thread processed which appear to be associated with it. Every day or so, one of these processes bursts into life and takes 100% of one of the CPUs for several minutes. The server is part of a two node cluster. It then goes back to normal. Does anyone know why?

# ps -ef|grep native
root 1383 1 0 Feb 7 ? 1315:59 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa
root 7961 1 0 Feb 17 ? 0:00 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa
root 14077 1 0 Feb 8 ? 0:00 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa
root 22791 1 0 Mar 7 ? 0:00 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa
root 23685 20630 0 10:12:17 pts/tc 0:00 grep native
root 11820 1 0 Mar 29 ? 0:00 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa




$ ps -ef|grep native
root 1414 1 0 Dec 27 ? 1960:39 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa
root 8645 1 0 May 2 ? 0:00 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa
root 7320 1 0 Feb 22 ? 0:00 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa
root 10363 10362 0 May 11 ? 0:01 /opt/sanmgr/jre/bin/../bin/PA_RISC2.0/native_threads/java -Djav
root 3046 1 0 Feb 4 ? 0:00 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa
root 22009 1 0 Mar 27 ? 0:00 /opt/sanmgr/hostagent/../jre/bin/../bin/PA_RISC2.0/native_threa
troplive 14354 14127 0 10:14:17 pts/td 0:00 grep native

# cat dialevents.log

*** Start Script Output ***
Date: 05.22.07 00:07:46

--entering DialEvents ...
getting HostInfo
*** xml ***


HostName="tripsb"
HostIpAddress="10.1.100.2"
LastIOChange="1179788864000"
OsType="HP-UX"
OsVersion="B.11.11"
HostModel="PA_RISC"
HostSerialNum="762957635"
JavaVersion="1.3.1.05"
JavaVendor="Hewlett-Packard Co."
JavaHome="/opt/sanmgr/jre"
JavaClassVersion="47.0"
JavaClassPath="./classes/dial_if.jar::./classes/jcore.jar:"
Version="03.00.00.0307"
DiscovPollPeriod="19"
StatusPollPeriod="7">
IpAddress="10.1.100.2"
DnsName="tripsb">


Version="03.00.00.0307"
Build="Release">

--entering DialEvents.getConnection()
login successfully..
--exiting DialEvents.getConnection()
deleted data/dialevent.xml
--exiting DialEvents()
:*** End Script Output ***

[Steven E. Protter]

Shalom,

Could be i/o wait, but I'm more in favor of this being caused by bad san manager software.

Bad code can kill a system almost as fast as I can.

:-)

sEP