Invalid address to attach the shared memory segment (1173)

[Harry Yu]

hi UX-experts,

Our HP-UX (10.20) stopped respond over the weekend. I am not able to telnet into the server (blank screen, blinking cursor.) I hard booted it and everything comes back correctly. When I tried to restart our database (Progress MFG/Pro), an error message "Invalid address to attach the shared memory segment (1173)" is displayed follow with "There is no server for database /db1/db. (1423)". I need a quick help to determine what is wrong and need to get the database up and running as all the people here are waiting for it to do their work. Thanks many times!!!

UX newbie,

Harry

[RAC_1]

ipcs -moba

Identify all segments where NATTACH is zero. Remove those with ipcrm and check

Anil

[Harry Yu]

RAC,

I'm not quite sure what do you mean. Can you give me a bit more details? Thanks.

[RAC_1]

Read the man pages of both commands.
ipcs will show you the shared memory segments, who own them etc. Sometime it may happen that a segment is not deattched proprly. In that case you can remove it with ipcrm command.

You should also try to look if there are any databse processes rnnning or not?

Anil

[Harry Yu]

Hi Anil,

Now I see what you're saying. I saw one of the line with NATTH 03, so I removed it and the same problem presist. Below is all that I've left...

------------------------------------------------
IPC status from /dev/kmem as of Mon Jun 14 10:25:35 2004
T ID KEY MODE OWNER GROUP CREATOR CGROUP NATTCH
Shared Memory:
m 0 0x2f100002 --rw------- root sys root sys 42
m 2 0x4e0c0002 --rw-rw-rw- root root root root 23
m 3 0x412007b1 --rw-rw-rw- root root root root 23
------------------------------------------------

Need more help. Thanks!

[RAC_1]

What else is running ont his host? What is the setting of shmmax kernel parameter.

You can that with kmtune -lq shmmax. OR through SAM.

Seems that your progress database is not getting the enough shared memory. You can also download free toold shminfo (shred memory information tool) from
floowing link.

hprc.external.hp.com/sysadmin/programs/shminfo/ (login=contrib,pw=9unsupp8).

Anil

[Harry Yu]

There is no other service running on the server. Our HP-UX 9000 is dedicated to the Progress MFG/Pro database. According to our users, it was still working last night till 11pm then everything is unknown. Again, all we got this morning was blank screen when tried to connect.

I'm not sure what does the "kmtune" command does, but it's not in my server... It has been down for 3hrs now and I'm getting desperate... :(

[RAC_1]

A plain reboot (of system)should get you started.
Once system reboots, try starting Progress. Post.

Anil

[Harry Yu]

Anil,

I restarted the server with both soft and hard boot and it doesn't do the trick. The problem could lie on the Progress side, that I'm not sure. Are you familiar with Progress? When I tried to start the production database, I received the follow errors...

Invalid address to attach the shared memory segment (1173)
There is no server for database /db1/db (1423)
** The server terminated with ext code 2. (800)

Any idea?

[RAC_1]

Not familiar with Progress. But I strongly doubt that this related to Shared memory problem.

Post following outputs. (run these commands as root)

echo "shmmax/D"|adb -k /stand/vmunix /dev/kmem
ipcs -mobp

Anil

[Harry Yu]

Anil,

That doesn't work either. The server did take the first command you gave me. I started to think it seems like the database is haven't a problem starting and not the UNIX... but thanks for your help anyway. If you got any other idea, please let me know. :)

- Harry

[RAC_1]

I am egar to know, what exactly is going wrong here. Did you login as root and execute the following commands?? Login as root.

/use/bin/echo "shmmax/D"|adb -k /stand/vmunix /dev/kmem
/usr/bin/ipcs -mobp

Anil

[Hein van den Heuvel]

This is NOT a shm problem.
Do NOT turn those kmtune knobs!
If those settings where fine last night, they are still good enough today.
This is assuming the DB configuration was not changed. Was it?

What else was not changed :-).
Seriously, any recent changes? More buffers? more files?

The messages strongly suggest that the db crashed and had a problem coming back up.
Do you have a db stop / start procedure?
Run it manually and see what gives.

Do you know anything about the db structure? Are it's files still there?

Do a find /db1/db -atime +1 to find out what what touched recently?
Any "*log" file modifoed recently?

Do you have right to MFG/Pro support? call them!

Good luck,
Hein.

[Harry Yu]

hi guys,

It turns out to be the Progress problem. We had QAD (our database company) fixed the problem. They've time-stamp problem when we use OSBackup instead of their own backup utility (call Probackup). The problem doesn't lie on the UNIX side. No wonder we can't figure it out. :) Thanks, guys.


- Harry