Unidata and Serviceguard question

[Bryan King_2]

I have a situation where Unidata 6.0 and Datatel in an MC/Serviceguard 2 node HP-UX 11i cluster are being started through a package. The cluster consists of rp4440 servers.

When Unidata is started up outside of MC/Serviceguard, Datatel startups properly. I tested this by taking all of the startup commands and putting them into a separate script, then running it from root.

When Unidata is started up through MC/Serviceguard, Datatel does not startup correctly. The datatel dmi process dies immediately.

The error messages are written to a datatel logfile.

The requirements of Unidata are that it be started by root and that two environmental variables are set. All three of these things are happening through Serviceguard.

Datatel can be started up by root or by a user named datatel

I looked at the Unidata process errorlogfiles (cleanupupd, smm, sbcs, udt) and when Unidata is started up through Serviceguard, they report an error "tty: no such device or address".

Any ideas or thoughts as to why this set of circumstances where Unidata starting up in Serviceguard would cause datatel not to start properly would be greatly appreciated.

[Sanjay_6]

Hi Bryan,

Looks like your process needs a valid tty to start. Can you try this in your startup command,

su - root -c". /.profile;your_script_here"

You set the env variable in the root .profile or through some other such script.

When you start the package through SG, there is no valid tty that is open and this is a usual error message.

you can also use the export TERM=vt100 or something in your profile and comment the interactive TERM setting if it is done through .profile

Hope this helps.

Regds

[Bryan King_2]

I solved the problem by issuing an su - root in the package control script before running one of the startup commands.

The problem turned out to be that Unidata expects a tty device be attached to the process that is running the startup.