unable to run package in SG cluster

[Neil Armstrong]

Hi all,

I've got a 2 server SG cluster that runs 2 packages, with one service defined in each.

One package starts IP-based client services, and works as expected (most of the time!).

The other starts X.25-based clients, and always fails - the log entry in the cluster script log simply states the package starting, then another entry (exactly the same time) stating that the package is being halted. No other log messages here or in syslog.

After this, a cmviewcl will show the package as Halted, Unowned but with switching Enabled. However, issuing a cmrunpkg command only gives this:

# cmrunpkg pgclldv_x25
cmrunpkg : Node pgcl1ldv is currently unable to run package pgclldv_x25.
Check the syslog on node pgcl1ldv and pkg log files for more detailed information.

No information in either logfile. Issuing a cmmodpkg -e pgclldv_x25 doesn't help either.

Any ideas? Are there some hidden flagfiles or somesuch that are out of date and in the way? Both nodes in the cluster have been rebooted a couple of times but still with the same issue.

thanks,
neil.

[Rajeev Shukla]

Hi Neil,

Can you post the syslog entries and package script log files for more information.

Cheers
Rajeev

[Hoang Chi Cong_1]

Hi

It seems that there is something wrong in the package configuration file.
Please attach the package log file and syslog.log file.

Regard,
HoangChiCong

[melvyn burnard]

wel the first question is:
Has it ever worked ok?
If yes, then what has changed?
If no, then it would appear that there is a configuration issue.
The probable reason for the message after issuing the cmrunpkg command is that switching has been disabled for thos package.
have a look at the result of doing:
cmviewcl -v -p pgclldv_x25
Pay attention to the switching settings.
Also which version of Servieguard and which patch are you using?
Do:
what /usr/lbin/cmcld

[Stephen Doud]

Please give us a copy of the package configuration file, package control script, and package control script log file from the server where the package failed to start.

Most immediate package shutdowns are caused by a trigger, most likely a SERVICE_CMD exitting with a 1 (see syslog.log).
Such a termination indicates that the monitor script launched by the SERVICE_CMD exitted because a monitored process was not in existence.
I am not familiar with X.25 packages, so I don't know if this is the case. However, syslog.log should indicate if a package service exitted with a 1 condition... and if that is true, inspect the SERVICE_CMD to see what it launches and why it's dependency would exit.