MC/SG rolling upgrade

[jh_yang]

Got a problem when I tried to convert old version cluster binary file to the new verion.
< unable="" to="" open="" local="" cluster="" handle.="" no="" such="" file="" or="" directory.="">
Anybody know what's cluster handle?
Also got an entry in syslog.log


Need help.
Thx
David

[melvyn burnard]

How about some more info, like what OS and SG versions involved, steps taken so far etc?

[jh_yang]

2 nodes cluster.
before upgrade: OS 10.20, SG: 10.11
after upgrade: OS 11.00, SG: 11.14

Step1: Switch the pkg to Node B
Step2: Upgrade the Node A to OS 11.00 and SG 11.14.
Step3: Restore the cluster files to Node A
Step4: Convert the old binary file on Node A

[melvyn burnard]

When you say:
Restore cluster files to node A

why did you need to do htis, if you have done an upgrade, hten teh files would still be htere

[Glen Liu]

I am working with David on this project. Actually we were not upgrading the systems. We did cool install on them and then restore the configuration from the backups. That's why David mentioned restoring pkg file.

We'd solved this problem by modifying the /var/adm/inetd.sec file. Originally, the following entry existed in the inetd.sec:
hacl-cfg allow NodeA NodeB

We changed it into:
hacl-cfg allow NodeA NodeB 127.0.0.1

And then the convertion completed successfully.

What we still don't understand is why the 127.0.0.1 IP was not resolved as localhost or loopback but unknown by inetd according the error message from syslog.

For the reason why it use 127.0.0.1 other than the primary IP address of the server, I guess it's because the convert utillity wants to must sure it's converting the local content. Can anybody confirm it?

Thanks!

[melvyn burnard]

well yes it does need that localhost.

I am glad you fixed it, but have to alert you to the fact that in fact the method you have used to do a "rolling upgrade" is not actually suported.