We have an ALE_Server adapter configured that correctly registers with SAP and everything functions OK, for a while. Then one of our SAP guys will call and report that the adapter is not registered within SAP, nor are any documents flowing. In adapter configurator everything looks OK and a test is successfull and has the effect of registering the adapter in SAP and re-starting the flow of documents.

To me it seems like SAP is resetting the IP connection and that WM is not recognizing this condition, or that WM is losing the connection without realizing it.

Has anybody else encountered this ? If so, how was it resolved ? If not, any ideas what I should be looking for ?

(Though I only speak of a single occurence above, this happens frequently on many different SAP systems within the organization)

Not sure if this is quite the same but I know the standard db adapters could not detect when the db cleared its connections. This was with DB2, where all connections were closed nightly. The db adapter had to be restarted when this occurred, so we scheduled a cron job to bounce the db adapters each day. I’m not sure if the intelligent adapters have the same issue but I’d guess they probably do. I don’t think the adapters can do anything about it, short of constantly pinging their resource.

I don’t know enough about SAP to say for sure, but speaking IP I believe that the adapter must connect to SAP as a client. This would mean that for the ALE_Server it SAP would have to send the request on the already established. Assuming this is not possible, then the connection had been lost earlier. Shouldn’t wM have received a CONNRESET ?

Otherwise if SAP is cycled wM will become confused.

That last post got a bit hurried and garbled due to local issues here

What I meant to say was that if wM is connecting to SAP as an IP client, then the connection can be lost in one two ways - either wM will close it, in which case it should know to open it the next time it is required, or SAP can close it, but this should raise a CONNRESET (shouldn’t it ?) informing wM that it will need to recreate the connection.

This does not seem to be happening and since I believe the CONNRESET is raised by the IP Stack I cannot see how SAP can be at fault here, so is wM the culprit ? If so, who else has encountered this ? and how was it addressed ?