AUTO_START in MC/SG 11.12

[Patrik Nilsson_1]

Hi,
I have a 2 node cluster where I cannot get my packages to auto start when I boot the servers. The cluster is started but my packages are not.
I have put "AUTO_START YES" in package config but after boot cmviewcl shows "AUTO_RUN disabled"

Here is my package configuration from cmscancl
#######
Package information:

package ID 24579:
package name: dbprogr
package global flags: 5
(Package Switch Enabled)
(Package Local Switch Enabled)
(Configured Node Failover)
(Manual Failback)
package run script: /etc/cmcluster/dbprogr/dbprogr.sh
package run timeout: 300 (seconds)
package halt script: /etc/cmcluster/dbprogr/dbprogr.sh
package halt timeout: 300 (seconds)
package primary node: progr
package alternate node: roam
package subnet: 192.168.1.0

package services:
service ID: 2
service name: dbprogr_serv
service halt timeout: 290 (seconds)
service fail fast: Disabled

package ID 57602:
package name: dbroam
package global flags: 5
(Package Switch Enabled)
(Package Local Switch Enabled)
(Configured Node Failover)
(Manual Failback)
package run script: /etc/cmcluster/dbroam/dbroam.sh
package run timeout: 300 (seconds)
package halt script: /etc/cmcluster/dbroam/dbroam.sh
package halt timeout: 300 (seconds)
package primary node: roam
package alternate node: progr
package subnet: 192.168.1.0

package services:
service ID: 2
service name: dbroam_serv
service halt timeout: 290 (seconds)
service fail fast: Disabled
#####

Any ideas? What have I forgot to do?
/BR

[Sridhar Bhaskarla]

Hi,

I have A.11.14 installed. The corresponding parameter is "AUTO_RUN YES" in the package configuration file. In older versions (A.11.09) this parameter was PKG_SWITCHING_ENABLED.

Attach your package configuration file if you can with dummy entries for the hostnames/subnets etc.,

-Sri

[Michael Steele_2]

Hmmm, my notes indicate 'PKG_SWITCHING_ENABLED' obsoleted in MC/SG version A.11.12, its indicated here.

From both nodes can you:

swlist -l fileset -a state | grep -i serviceguard

-and-

swlist -l fileset | wc -l

Also, did you verify the package?

cmcheckconf -P configuration.file

[Geoff Wild]

IMHO, I would never have MC/SG nor it's packages auto start on a reboot. I don't do that now.

If a server crashed/rebooted on it's own - you should determine "why" prior to starting up any packages...for example, what if the server that has your Production system crashes? well, Prod would then start up on your fail-over node...then, when your crashed box starts up, Prod is going to try to start - it should fail and will mark the package as non startable on that node...

What if a disk was failing on the Prod box - and when the machine rebooted - it was worse - then you may corrupt your database when the package starts up....

Just something to think about...

Rgds...Geoff

[Patrik Nilsson_1]

Hi,
I'm sorry to confuse you but I misstankingly wrote that I'm "AUTO_START YES", instead I'm using (the correct parameter) "AUTO_RUN YES" in my config file.

I have attached both my cluster and package files to this message.

The output on both nodes of:
swlist -l fileset -a state | grep -i serviceguard

# ServiceGuard
ServiceGuard.CM-SG configured

/BR

[Sridhar Bhaskarla]

Hi Patrick,

Well in that case, I would suggeset to reapply the package configuration and observe 'cmviewcl -v -p package_name".

-Sri

[melvyn burnard]

So what do the package logs say?
if the packages have attempted to start at all, it will be logged there, and probably the reason why it fails to start.
My guess is it tried to start on one node, fails, tries to start on the other node,, and then fails again.Also, what does cmviewcl show?

[Michael Steele_2]

A) Do you have AUTOSTART_CMCLD=1 in /etc/rc.config.d/cmcluster?

B) In the cluster.ascii file there is also the parameter 'AUTO_START_TIMEOUT'. Default 600 seconds. Compare this number after using this formula.

"...value equal to the boot time of the slowest booting node minus the boot time of the fastest booting node plus 600 seconds..."

Beside checking for package errors in 'cmcheckconf -P' run 'cmscancl' and attach.

Finally, there is node enabling as well as package switching:

cmmodpkg -e package
cmodepkg -e -n node