Problem with Informix in MCSG

[yc_2]

Hi,

We have just setup the following 3 weeks ago:

Building A:
(A.1) rp4440 (HP-UX 11.11, MCSG and Informix)
(A.2) EVA3000 (mirrored Informix data to (B.2)

Building B:
(B.1) rp4440 (HP-UX 11.11, MCSG and Oracle Financial App)
(B.2) EVA3000 (mirrored Oracle data to A.2)

The environment was running fine till last Friday.

An accidental power re-cycle (not in proper sequence) of the equipment in Building A had caused the EVA not to be recognised by the server.

The problem was rectified after a proper power recycle but the physical volume did not match with the lvmtab. vgexport and vgimport were done to rectify the problem.

Today, new problem surface. The Informix package is not able to startup (intermittently) after it is halted. Sometimes after a few attempts (using cmrunpkg or cmmodpkg), then the package starts up. When the CNTL log file and syslog file were examined, sometimes no entry is in the logging file and sometimes the following statement display:

oninit: Fatal error in shared memory initialization

ERROR: Failed to start Informix server informix with /informix_runt/bin/

oninit.

PS: The shared memory is cleared (ie no shared memory ID is owned by Informix) before the package is started.

When the package is successfully started up, the secondary node switching is disabled.


Any pointer is appreciated.




Rgds,
YC

[RAC_1]

Looking at the message - "oninit: Fatal error in shared memory initialization", it seems that informix is having problem getting the required shared memory area. Before you start, look at ipsc -moba, Check how much is free and how much is used. There is also a tool as shminfo (shared memory information)

What setting in Informix controls this?? (I mean sometinng as SGA in oracle)

Anil

[John Poff]

Hi,

Maybe when Informix comes down it isn't getting shut down cleanly and is leaving some shared memory segments open, which causes it problems when it tries to startup. I've seen similar problems with Oracle before.

JP

[Rainer von Bongartz]

Please post the relevant parts of your online log and your $ONCONFIG.


Regards
Rainer

[Steve Lewis]

As you already said that you made sure the shared memory segments for that instance are cleared, it should not be a problem with that.

The next most likely candidate is the port it is listening on. If this were the case it would say 'Cannot bind a name to the port' in the informix message file.

If the INFORMIXSERVER is bound to the IP of the PACKAGE in sqlhosts, not the server, if you try to start it when the MCSG package is down, then it would not work. Similarly, if your informix installation is outside of the MCSG package and you attempt to bring up the database when the package is on the other node, then you will get this message.

If you post the output of onstat -m when it fails to come up, it would help us diagnose more fully where the problem lies.

[yc_2]

Sorry for the late reply.

For ipcs -moba:
- How to know how much is free and how much is used?
- Can't find shminfo

echo $ONCONFIG
onconfig.prd

I'm not familar with Informix, what is the function of INFORMIXSERVER?

The problem is intermittent, I will post the onstat -m output when the pkg not able to startup.

[oczkov]

I guess it is to late for the solution, but we have kind of similiar situation (electricity problems).
The solution could be like that:

(1)Enter your /your_dir/informix/etc/
(2) Move .conf.server_name.infos and oncfg_server_name files to some other directory (/home/abc)
(3) Remove /tmp contents. IDS seems to put some files or dirs in there.
(4) Try to start Informix again.

In our case it helped. IDS recreated the files moved in step 2 and everything started to work.

Regards.

[yc_2]

Found certain databases corrupted. Restored back the databases and everything back to normal.

Thanks you everyone for the effort you have put in.