problem navisphere agent hpux

[Gerrit_1]

Hi,
I'm having a problem with the navisphere software.
This is the situatie :
a production-host is connected to a cx600 and
a cx300.

Version of the navisphere-agent :
root/sv00247#swlist -l product | grep -i navi
NAVIAGENT 6.19.1.3.0 Navisphere Disk Array Management Tool (AGENT)
NAVICLI 6.19.1.3.0 Navisphere Disk Array Management Tool (CLI)

The config-file :
clarDescr Navisphere Agent
clarContact Tom Geudens, 2380

device auto auto

user root
user root@sv00247
user system@10.101.0.16
user system@10.101.0.17
user system@10.101.1.17
user system@10.101.1.18
user system@10.102.0.17
user system@10.102.0.18
user system@10.102.4.22
user system@10.102.4.23

# OptionsSupported AutoTrespass

poll 120
nomegapoll
eventlog 100
baud 9600

The navisphere-agent start, but after 120seconds, it generates a core- file.

root/sv00247#file /core
/core: core file from 'naviagent' - received SIGSEGV

Any ideas ?

[Chan 007]

Hi,

Did you check your agent.log. It is in /etc/log. Also check the HostFile.txt. This should have your systems (host name and/or ip address only)

Change this and restart the Agent once again.

Chan

[Mark Fisher_4]

Guess what? I have the exact same issue. naviagent 6.16 will work just fine, but 6.19 will always crash as soon as the polling interval expires. change your "poll" parameter from 120 to 60 and watch it crash in 60 seconds.

The problem is a failure in the lseek call to update the navimon.log file. We have a core dump and tusc/truss output:

[14802] write(10, "01\0\0 \0c6d 1a180201\0\0\0\f\0".., 512) = 512
[14802] lseek(10, 0, SEEK_SET) ........................... = 0
[14802] write(10, "\0\0\014\0\0b 14\0\0\014\0\0\0\0", 16) = 16
[14802] lseek(10, 25108, SEEK_SET) ....................... = 25108
[14802] Received signal 11, SIGSEGV, in user mode, [caught], partial siginfo
[14802] Siginfo: si_code: SEGV_MAPERR, faulting address: 0x60000000c31f8d68,
si_errno: 0


I have reported this to EMC and they have an engineer working on it. I would recommend logging a call with EMC as well to get the pressure on.

But, forthe HP/UX guru's. Why does this simple lseek call fail? They issue a write to the file, then lseek with offset zero to reset the pointer, issue another write, then lseek again with the updated pointer from the write and lseek gets a SIGV_MAPERR. The updated pointer looks good and points to the end of the file.

Naviagent 6.16 does the exact same sequence of calls and does not fail. Plus this whole thing is a PA-RISC image and is running under the aries emulator. And after the lseek fails, every other system call like time also gets SIGV_MAPERR. This repeats over and over until the emulator runs out of memory and you get the core:

ARIES32 Limitation/Error:

PID 10367 received SIGSEGV for stack growth failure.
Possible causes - insufficient memory or swap space,
or PA stack size exceeded pa_maxssiz_32bit.

I increased pa_maxssiz_32bit and created a .ariesrc file with the entry:

/ -ssz 320000

This does not prevent the issue, not should it either.

So is this an EMC problem or a HP/UX issue?

[Arijit]

The signal is not due to the failure of the lseek() system call. The return value of the system call is not an error.
There seems to be an access to the address 0xc31f8d68 after that which faults.

When you do get a core file , is it just called 'core'?. Any core created by ARIES should be named core..

Also, make sure you have the latest ARIES patch installed, PHSS_34201 (downloadable from ITRC).
For more information on ARIES you can visit the site www.hp.com/go/aries

[Mark Fisher_4]

We actually figured this one out. In the core we were also crashing in a setenv call. It was discovered that we had an incorrect SHLIB_PATH entry. The following 2 entries were removed from SHLIB_PATH and naviagent 6.19 does not crash any more:

/usr/lib and /opt/Navisphere/lib

These entries were in the root .profile file. They were there because the Navisphere install guide instructed us to put them in there. Now it appears EMC does not want them anymore.

[Sยภเl Kย๓คг]

Hi Gerrit,

Could u please tell me, how this issue is resiolved.. what was the solution for this..

I have the same problem..

Thanks, well in advace.

[Klaus Page]

Thank you Mark Fisher. Problem resolved by removing the 2 entries from SHLIB_PATH.
We had exactly the same problem and it occured only after we re-installed the O/S. We installed HP-UX 11i Version 2( Media Kit released December 2006 ).

[Basheer_2]

Hi Gerrit,

These are my versions on HPUX 11.11 and work fine for me.

NAVIAGENT 6.19.4.7.0 Navisphere Disk Array Management Tool (AGENT)
NAVICLI 6.19.4.7.0 Navisphere Disk Array Management Tool (CLI)