08-09-2013 12:34 PM - edited 08-09-2013 01:29 PM
We are moving a CMS library between a Alpha and Integrity server. Both servers are running OpenVMS 8.4 and CMS version 4.5. Prior to the transfer "CMS Verify" was run to check the status of the library. The library then transferred using backup, and restore on the Integrity server We again run "CMS Verify" run the Integrity server to check the status. As a final check we did a simple "cms show history" then we noticed the history was not same. We have run difference between the "*.HIS" files between both servers and no difference were found. The number and files size were check and no issues were found. Finally ran "CMS SHOW library/FULL" on both servers and again no difference were found.
More detail checks we have found that after the transfer CMS is not displying "Create", "Fetch", or "Verify" history records on the elements in the library.
Any ideas as to why we are see a difference in the history.
08-12-2013 09:06 AM
If I use "/trans=create" I get the "create" records plus all other history records. On the Alpha box that command gives me just the "create" records only".
The command "/trans=all" does not show all the transactions either.
If I use "/trans=create" it looks like the tranfser between the servers is correct, it just that CMS on the Integrity server has issue with displaying the records correctly.
08-13-2013 01:59 AM
Looks like a bug in CMS for I64. There are more commands which aren't listed with SHOW HISTORY, for example INSERT. The bug also shows in CMS libraries created on I64. This affects my tool which uses the history output to create a git repository (on a non-VMS system) from a CMS library.
There seems to be a workaround: SHOW HISTORY /NOTRANSACTIONS.
Did you report this to HP and did you get any feedback?
08-14-2013 03:37 PM
This says a lot about HP's testing.
Wouldn't you think that the correct functioning of CMS History would be rather important to users?
08-20-2013 01:13 PM
I'll guess (but I don't know) that a CMS test system involves creating a new library, doing all sorts of commands, and checking the history. Testing the ability to "import" a history file from another system is less common and perhaps not tested as well. I'm surprised that the format is different. I wonder if it works correctly in a mixed-architecture cluster with shared storage?