Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Matt Doidge said:
> https://ggus.eu/ws/ticket_info.php?ticket=94096 (15/5) RHUL were ticketed
> over publishing the obsolete "GLUE2EntityCreationTime". Am I the only one
> to find tickets like this a little confusing? Govind heroically fought the good
> fight, and this ticket can now be closed. In Progress- can be closed (22/5)
I'll try to explain as it may happen to other sites. GLUE 2 objects have a creation timestamp, and for objects from dynamic providers an old timestamp (more than say an hour or so) indicates a problem somewhere - however there are some CE objects taken from static LDIF files which may obviously have rather old times.
Unfortunately there's a bug in the current BDII, which means that if an object stops being published it doesn't get deleted, it just hangs around with whatever timestamp it had last (GLUE 1 is OK, this is just for GLUE 2). You can fix that by restarting the site and/or resource BDII, ideally after deleting the cached list of URLs it queries (/var/lib/bdii/gip/cache/gip/site-urls.conf-glue2) - I think that's only relevant if a whole node has changed but I don't swear to it. This is e.g. why some sites were still failing the EMI 1 test after they upgraded.
Usually objects don't vanish very often, but unfortunately*2 there was a bug in CREAM which meant that software version tags were converted to lower case. When you upgrade to the latest CREAM the tags all go to the correct case, but due to the first bug all the (many) objects for the old lowercase tags may stick around. That doesn't do a huge amount of harm, but it's still good to clean them out, hence the tickets for sites which had a large number of dead objects - there were only about half a dozen sites like that.
Stephen
--
Scanned by iCritical.
|