- SAP Community
- Products and Technology
- Additional Blogs by SAP
- Hints about PI runtime cache refresh
Additional Blogs by SAP
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for
former_member67
Discoverer
Options
- Subscribe to RSS Feed
- Mark as New
- Mark as Read
- Bookmark
- Subscribe
- Printer Friendly Page
- Report Inappropriate Content
03-09-2010
9:15 PM
This blog provides some useful hints about XI/PI runtime cache refresh from the following three aspects in order to avoid the interruption on message processing caused by cache refresh.
- When XI/PI runtime caches need to be refreshed or cleared
- The potential risk when perform cache refresh
- The proper way to perform cache refresh
In the first chapter, XI/PI runtime cache overview is introduced via a summery sheet and two diagrams. In the third chapter, the documents related XI/PI runtime caches are provided.
1 XI/PI Runtime Cache Overview
The following summary sheet and diagrams provide the overview of XI/PI runtime caches.
2.1 When perform cache refresh manually
There are three types of cache refresh on XI/PI critical runtime caches "IS ABAP cache", IS Java/Mapping Runtime cache" and Adapter Engine CPA Cache":
- MAPPING: Interface/operation mapping
- XI_TRAFO: Message mapping
- TRAFO_JAR: Imported archive
- RepBProcess: Integration process
- AdapteMetaData: Adapter metadata
The cache refresh will cause failure of
- All messages under processing
- When Integration Engine RUNTIME parameter CACHE_DIRTY_READ set as 0, which will terminates message processing if a pipeline service is used and the cache refresh updates objects for this specific service. In worse case, queues in SMQ2 can be stopped with error text like "Queue stopped due to cache update", which will block message processing at Integration Engine and needs to unlock stopped queues in SMQ2 manually.
- When full/complete cache update failed on cache instances
- Some messages under processing
- When Integration Engine RUNTIME parameter CACHE_DIRTY_READ set as 1 (default value), which will not interrupt message processing during cache refresh. This is achieved by simply using the status of cache before the cache refresh for the evaluation. Whether this results in outdated data being used depends on whether the messages concerned are in any way connected with the changed data. However, some of the objects may not yet be available in the XI cache during cache refresh and configuration errors may occur for the messages.
- When delta/partial cache update failed on cache instances
The failed asynchronous messages can be restarted from SXMB_MONI manually or by the job running report RSXMB_RESTART_MESSAGES automatically. The failed synchronous messages must be resend from sender systems. For detailed information, please refer to note 1337797.
Do not perform cache refresh when there are messages under processing, especially there are synchronous messages under processing.*Partial/Delta cache refresh is preferred, which is enough to update all changed IR and ID objects as well as to solve the issues of missing or incorrect version of ID/IR objects in caches. </p><p>Complete/full cache update is required only in case of emergency when there is inconsistency in PI runtime caches. Please grant the authorization for complete/full cache update only to PI system administrator according to note 1122879.*
h5. 2.1.2 Things to do when perform cache refresh manually
To avoid cache refresh failed, please follow steps 1, 2, 3, 4, 5 and 6 before performing XI/PI cache refresh manually when necessary.
1. Run "cache connectivity test" via RWB Component Monitoring first
2. Check the setting of the Integration Engine RUNTIME parameter CACHE_DIRTY_READ via transaction SXMB_ADM/SXMB_ADMIN. CACHE_DIRTY_READ=1 is the default and preferred setting.
3. Please consider to increase the value for the following parameters when there are large number of ID objects (for example: more than thousand) needs to be updated to XI/PI runtime caches in order to avoid cache fresh failed due to timeout error or http_communication_failure.
- JVM parameter: -Dorg.w3c.www.protocol.http.connections.timeout=600000. This parameter is valid for performing CPA cache refresh via https connection. For detailed information, please refer to note 1365480.
- The properties "ServletInputStreamTimeout" and "ServletLongDataTransferTimeout"of J2EE http service via NetWeaver Administration Page (Only valid for XI 3.0 & PI 7.0, but not for PI 7.1). For detailed information, please refer to note 807000.
- The timeout option of RFC destination INTEGRATION_DIRECTORY_HMI via transaction SM59
4. For XI/PI Integration Server system or XI/PI non-central adapter engine system consisting of the central and dialog instances, please check the following properties of CPA Cache service com.sap.aii.af.cpa.svc are defined using the hostname and port number of Web dispatcher or hardware load balancer.
- SLD.selfregistration.hostName
- SLD.selfregistration.httpPort
- SLD.selfregistration.httpsPort
5. Check the user who perform cache refresh has enough authorization.
- Having role SAP_XI_APPL_SERV_USER and the authorization S_RFC_ADM is contained in this role. For detailed information, please refer to note 1177441.
- Having role SAP_XI_ADMINISTRATORABAP and the authorization S_XMB_AUTH is contained in this role, which is required for "Full/Complete Cache Refresh" of PI 7.1. For detailed information, please refer to note 1122879.
6. Check the following XI/PI system users are not locked via transaction SU01.
|
h5. 2.2.2 Things to do after ID/IR objects transportation
- Check the cache notification and cache update status in RWB Cache Monitoring and make sure both cache notification status and update status have green lights on all type of caches (ABAP, JAVA, and Adapter Engine Cache). If the status of any cache notification or cache update is flag with red color, please follow "How to handle XI/PI cache" guide as well as XI/PI troubleshooting guide to solve the issues during cache notification and update.
- For Web Service scenarios in PI 7.1, the Web Service RT Configuration needs to be checked via WSTASKCHECK in PI system and ABAP sender system to make sure there is no pending SOAP RT configuration task. If there is any pending configuration tasks, please schedule job SAP_SRT_CFG_PENDING_TASKS via WSTASKCHECK. For detailed info, please refer to note 1259998.
- Restore business processes in sender / receiver systems
2.3.1 Risk & Warning
*SLD Cache Location </p></td><td colspan="2" width="500" valign="bottom"><p align="center">Clear SLD Cache</p></td></tr><tr><td width="248" valign="bottom"><p align="center">XI 3.0 / PI 7.0</p></td><td width="252" valign="bottom"><p align="center">PI 7.1 and above*
ABAP stack | 1. Run transaction SE37 and execute function LCR_CLEAR_CACHE or | |
Java stack | Transaction SXMB_IFR -> Administration Link -> Runtime Tab -> Cache Overview -> Click "Refresh Cache" button on SLD Cache | Transaction SXMB_IFR -> Administration Link -> Runtime Tab -> Virtual Machine Cache Overview, Select SLD Cache and Click "Refresh Selected Caches" button |
Adapter Engine cache store the URLs of Adapter Engines including the central and non-central Adapter Engines, which is used by the Integration Engine to pass a message to the central or non-central Adapter Engine messaging system when the receiving system uses a JCA based adapter channels.
h5. 2.5.1 Risk & WarningWithout clear Adapter Engine cache upon the change to Adapter Engine URL, this may cause:
- The failure of message processing when the hostname and port number of Adapter Engine URL has been changed
- Uneven load distribution across multiple Java instances when the Adapter Engine URL has been changed to point to the hostname and port number of a Web Dispatcher or a hardware load balancer
To remove the changed Adapter Engine URL from Adapter Engine cache, please run transaction SXI_CACHE, goto menu "AE Cache" and delete the cache content of AE URL. The AE URL will be loaded automatically to AE cache when the first time a message requires it.
h4. 2.6 RFC Adapter CacheThe RFC Adapter has a cache for the metadata of function modules. This means it caches the definition of function modules, structures and other data types.
When the particular data is needed during runtime first time, the cache is filled from the configured RFC metadata repository. Once if a particular metadata has entered the cache, all the subsequent references to this function module are directed to the cache and no lookups to the RFC metadata repository are made for this type of metadata. Each communication channel has its own cache for RFC function metadata.
When a function module is changed in the backend, the changes are not published to the runtime RFC adapter cache. For detailed information, please refer to note 730870.
h5. 2.6.1 Risk & WarningWhen the import/export list of a RFC function is changed in the backend system, the RFC function metadata will be out of date, which will cause:
- The failure of message processing in XI/PI system or
- The runtime error of RFC function at the backend system
- Reload the RFC function module metadata into IR/ESR.
- Change the mapping program correspondingly and activate the change.
- Check the cache notification and cache update status in RWB Cache Monitoring and make sure both cache notification status and update status have green lights on all type of caches (ABAP, JAVA, and Adapter Engine Cache). If the status of any cache notification or cache update is flag with red color, please follow "How to handle XI/PI cache" guide as well as XI/PI troubleshooting guide to solve the issues during cache notification and update.
- Clear the RFC adapter cache by restart the related RFC channel in RWB channel monitoring. Restart RFC adapter service "com.sap.aii.adapter.rfc.svc" in NWA can also clear all RFC adapter caches, but it will affect all RFC channels.
IDoc adapter cache stores IDoc metadata at runtime as soon as a message needs this metadata for the first time. There is no automatic refresh when metadata is changed. IDoc adapter cache needs to be cleared manually via IDX2 or report IDX_RESET_METADATA.
h5. 2.7.1 Risk & WarningWhen IDoc metadata in sender or receiver systems are changed either via upgrade or development, the IDoc metadata in IDoc adapter cache will be out of data, which will result in:
- IDoc messages from a sending system may fail during converting IDoc to XML at IDoc adapter in XI/PI system
- IDoc messages to a receiving system may fail at inbound posting in a receiving system
For detailed information, please refer to note 907957 and 1157385.
h5. 2.7.2 Things to do when IDoc metadata changed in the backend systems- Import IDoc metadata into IR/ESR
- Change the mapping program correspondingly and activate the change
- Check the cache notification and cache update status in RWB Cache Monitoring and make sure both cache notification status and update status have green lights on all type of caches (ABAP, JAVA, and Adapter Engine Cache). If the status of any cache notification or cache update is flag with red color, please follow "How to handle XI/PI cache" guide as well as XI/PI troubleshooting guide to solve the issues during cache notification and update.
- Clear the old IDoc metadata from IDoc Adapter Cache via IDX2 or report IDX_RESET_METADATA
3.1 XI Directory Cache
3.1.1 XI Directory Cache in Detail
A very detailed documentation can be found at:
[https://wiki.wdf.sap.corp/wiki/pages/viewpage.action?pageId=10289567 | https://wiki.wdf.sap.corp/wiki/pages/viewpage.action?pageId=10289567]
h4. 3.2 How-To-Guide
How to handle caches in XI/PI will help you to
- To understand how data is entered and updated in caches in the Integration builder, the runtime and the adapters
- To know where caches are used
- To learn how to troubleshoot cache problems
h5. 3.2.2 How to Handle XPI Caches in SAP NetWeaver 7.0 (2004s)
http://www.sdn.sap.com/irj/sdn/go/portal/prtroot/docs/library/uuid/c0332b2a-eb97-2910-b6ba-dbe52a01b...</p>h4. 3.3 Troubleshooting Guide
3.3.1 Troubleshooting Guide - SAP XI 3.0/SAP NW 7.0, chapter 11 Caches
http://service.sap.com/instguidesNW04 ->Operation -> SAP XI
http://service.sap.com/instguidesNW70 -> Operation -> Process Integration
h5. 3.3.2 Troubleshooting Guide - SAP NetWeaver PI 7.1, chapter 10 Cacheshttp://service.sap.com/instguidesNWPI71 -> Operation -> Process Integration
h4. 3.4 Useful notes1383148 | Error in cache connectivity test |
1372866 | SAP_XI_ADMINISTRATOR added S_XMB_AUTH,value=CACHE, activity=01 |
1377033 | How to perform a Single Repository Object Cache Refresh |
1365480 | Error in full cache refresh in https mode |
1342064 | Insufficient monitoring for SXI_CACHE |
1337797 | XI cache: FAQ about parameter CACHE_DIRTY_READ |
1259998 | FAQ: Central Web service configuration using PI |
1177441 | XI cache: Error when reading HTTP destination |
1157385 | IDoc adapter: IDocs with status 03 and in XI not transferred |
1117249 | Incomplete Registration of PI components in SLD |
1122879 | Introduction of a restricted permission for cache refresh |
907957 | IDoc adapter: Metadata inconsistent after update |
897656 | NO_MAPPINGPROGRAM_FOUND after you import a SWCV |
881276 | CPADirectoryCacheException during J2EE Engine startup |
865137 | Full CPA Cache refresh can cause duplicate channel execution |
857890 | Cache status after deleting SWC or hierarchy relationship |
807000 | Http requests are not fully read after timeout |
764176 | Error in XI due to inconsistent SLD contents |
741214 | Troublesh. during cache update of the J2EE CPACache service |
732681 | XI 3.0: Buffering SLD data |
730870 | FAQ XI 3.0/ PI 7.0/ PI 7.1 RFC Adapter |
I would like to show my appreciation to the colleagues who support me and provide useful information during my blog creation.
- PI development colleagues: Frank Admas, Dimiter Dimitrov, Mandy Krimmel, Shibin S, and Nikhil Rao
- PI China CoE team colleagues: Jessie Han, Josef Schmidt and Tristan Hu
8 Comments
Related Content
- Personalization for tables in HTML Container for SAP UI5 application in Technology Q&A
- Configure Custom SAP IAS tenant with SAP BTP Kyma runtime environment in Technology Blogs by SAP
- Part 2:- SAP Basis OS support on SUSE Linux for beginners in Technology Blogs by Members
- ABAP RAP - Behavior Projection and Implementation Class in Technology Q&A
- Single Sign On to SAP Cloud Integration (CPI runtime) from an external Identity Provider in Technology Blogs by SAP
