Hi,
We got an unpredictable error on a workflow, after a moveDocument operation. The next operation is FetchDocument and sometimes we got the error "Caught error: org.nuxeo.ecm.core.api.ClientException: Failed to get document" ... (new path of the document). But the document exists at the indicated path. If you got back on the document and retry the operation, it can fail again or succeed. Recently, two users ask us to do something about this blocking exception and the only thing I found, is to do it again and again until it works.
Looking for this error in log, show me that this problem had happen times to times (but no one complain) but since June, it happens much more frequently (I don't think this can be correlated with an increase of activity, but I can still be wrong)
We got a nuxeo-5.8.0-HF31-1.0.0 with nuxeo DM.
More details on our workflow and automation chains :
We got two main "critic" steps when we perform a move document. The first, is to deals with the read authorisation of the workflow document and the second is to rename the document with some workflow data (users can found more easely documents).
When our workflow start, the support document is created on a temporary "workspace". The first task of the creator of this document is to indicate in which workspace this document had to be (the choose is made with a widget "single directory suggestion" to limit answer with user's authorised workspaces). When the "next" transition is call, and if the workflowVariable for the workspace is filled, an automation chain is launch with the following operations :
...
- set some Context Variables (document Id, Name..., workspace choose by user, and document path where we want put the document i.e. here directoryEnd)
- Fetch > Document (value : @{Context["doc"]})
- Document > Move (target : @{directoryEnd} ; name : @{docName})
- Fetch > Document (value : @{directoryEnd}/@{docName}) <- the one's which sometimes fails
If it's OK, the user can fill the some other variables of the workfow. Two of them are also used to rename the document at this second step. This other DocumentMove/Fetch operations can also fails sometimes.
Is it something bad in our automation chain ?
A colleague suggest that the fetchDocument operation can be performed before the ending of the Move operation. Is it a possible explanation ?
What to do to correct it ?
Thank for your help
