Problem with batch jobs

[Hoang Minh Tuan]

My system has two D390 servers in MC/ServiceGuard cluster, database is Oracle and Application use C language.
But it's a strange when it's not fork other processes when run batch jobs at the end of day while it run normally all day (processes at the end 's different from the day and this process has not yet connected to DB)
Application, Oracle DB run in Diskarray, only OS and Oracle software in /home of 2 servers.
When I switch to server2 (of 2 node) , it run normal. It's the same OS between 2 server.
I 've checked hardware, It's normal and dont have errors or warning log.
I've used make_recovery tape for 5 months ago to recovery OS, it's run normal only a week and it give out the same problem.
Has anyone could help me in this situation.

[Steven E. Protter]

You should be able to route the email the output of the make_tape_recovery job and get error output.


add

| mailx -s "Recovery Tape" yourname@yourdomain.net

to the cron command line.

I'd think about checking nfile and nproc kernel commands on your system.

SEP

[Hoang Minh Tuan]

I've compared kernel parameter between 2 servers, it's the same.
Restore server1 from make_recovery tape, it's run normally only a week then it come back the same failure before restoring.
I've checked hardware, it's OK.
Has anyone can give out other solution?
Thanks and regard