It's with great delight I read that the Solaris 10 Recommended and SunAlert patch clusters have been improved. Ed Clark has sat down and finally improved the install_cluster script, now called installcluster, and the cleaned up the directory structure. The whole build and verification process has been improved too.

The improvements are quite significant and well detailed, so I'll just summarize the improvements:

  • No more "false negatives" reported for things like patches that already been applied.
  • Immediate exit when a legitimate error is encountered, ie now one of the old false negatives from above.
  • New context intelligence for patching operations, thus providing better feedback of what is expected of the administrator.
  • Improved Live Upgrade integration. Ensure you read the cluster README for more details on this.
  • Improved filesystem space checking.
  • Improved logging. See the README for more details on this.
  • The patch_order file is now optimised for efficiency and to avoid known issues.
  • Cleaner directory structure

Just a few words of caution...

  • The cluster download is huge, so ensure you are up to date on your zip patches (see Doc 252447 for details and patches) before attempting to unzip it.
  • Patching as part of a jumpstart installation (using the "patch patch_order <path />" profile line) expects the patch_order file to be in the same directory as the patches themselves. This isn't the case, so until this bug is fixed, you'll need to remember to copy the patch_order file into the patches subdirectory.

All in all the improvements in both the cluster itself and the building and testing procedures look good and this should make administrators' (and support engineers' ;-) ) lives significantly easier when it comes to patching.