Hi all - I have just posted the following OSS message to SAP regarding some memory and file adapter problems we are having and was hoping you wonderful people could provide some advice too. We have a number of issues and I would really appreciate you having a read through and offer any advice. I have copied the whole message as it was sent to SAP.
Many thanks,
Stuart Richards
OSS Message:
We are having major problems with our messages and file adapters in XI 2.0 that we believe are all linked to a memory issue (possibly a memory leak) in our Productive system.
1. Outbound file adapters saying they’ve written data to text files but the data does not appear in the files. This is a major problem for us and our stores who are not getting invoiced for deliveries they are receiving. The file adapters in question are Prd_File_Out_SC_Halo_Advice_Note and Prd_File_Out_SC_Halo_Dts_Orders.
2. Message queues showing a SYSFAIL status in SMQR with an error message saying restart the LUW. This seems to occur when the system is at its busiest and is trying to process 1200 daily files incoming from our stores. These files are not large however and XI should easily be able to handle this. Restarting the queues or using RSXMB_RESTART_MESSAGES normally sorts this but it happens a lot of the time and for no good reason; there is nothing wrong with the data.
3. Sometimes the messages in error in point 2) above can only be restarted by restarting a relevant file adapter. We have many cases of file adapters that look like they are OK (green lights and STARTED status) but have ceased working. The end of the adapter log normally shows that some message is in the middle of processing when the adapter has just stopped working, despite it saying that it is OK.
4. File adapters are often unable to process incoming file and the log shows java.lang.OutOfMemoryError errors. The current file adapter doing this (Prd_File_In_DC_Stock_Balances) is trying to process a 5Mb file and it can’t. This means we cannot update SAP with our warehouse stock balances; a big problem and 5Mb is not a big file!
5. The main way in which we alleviate the memory error in point 4) above is to restart the adapter engine. All adapters are configured to start automatically when this happens but not all of them do. Of the 35 files adapters we have, it is quite normal for 2 or 3 of them (sometimes the same ones, sometimes different ones) to not start properly and sit green lit but in an INITIALIZED status. These have to be started manually.
6. General performance of messages going through XI 2.0 seems very poor. When we are at our busiest in XI, we have a couple of thousand messages being processed in queues. This is not that many, yet we see system utilisation of 95+% and very slow performance.
I have found OSS note 667711, which may be something we can try but we need further on whether this is relevant and how it can be applied if it is. There also seems to be a process you can see in ST06 that is using between 300Mb and 900Mb of memory at any one time; this seems excessive. The process is a JRE process and I think is related to the adapter engine and file adapters but am not sure.
Kind regards,
Stuart
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.