Details
-
Type:
Bug
-
Status:
Closed
-
Priority:
Critical
-
Resolution: Fixed
-
Affects Version/s: 1.4, 1.5
-
Fix Version/s: 1.5
-
Component/s: Taverna Core
-
Labels:None
Description
When Taverna starts up, and really whenever Taverna is used, it assigns LSIDs to all data allocations made, including workflows.
By our distributed default configuration, this is done by using the LSID provider at Phoebus:
taverna.lsid.providerclass = org.embl.ebi.escience.baclava.AssigningServiceClient
taverna.lsid.asclient.endpoint = http://phoebus.cs.man.ac.uk:8081/authority/services/AssigningWebService
taverna.lsid.asclient.ns.wfdefinition = operation
taverna.lsid.asclient.ns.wfinstance = experimentinstance
taverna.lsid.asclient.ns.datathingleaf = lsdocument
taverna.lsid.asclient.ns.datathingcollection = documentcollection
If this service on Phoebus, (currently located at the tomcat at /disk2/jakarta-tomcat-5.0.18 ) is down or slow for some reason, Taverna stops working, in some cases don't even respond or start up properly.
This means that if we turn off phoebus or something happens to be wrong with this single machine, including errors in the network routing between us and the random user, the random user will experience a non-functioning Taverna.
The LSID provider code should use the LSID endpoint with a rather low timeout (say 0.5 second) and failover to a local pseudo-random generator automatically, as a working Taverna is better than ensuring globally unique LSIDs.
Working on this now, hi-pri