TREX Setup for the Enterprise Search scenario
It has been noted that many customers are still running their TREX for the Enterprise Search scenario in a non-recommended manner. Specifically they still use either a Master-Slave arrangement or are not taking care of the TREX delta merges using CRON.
ESH manages the creation of all its indices in TREX. On an initial indexing the data is indexed to the main index directly. Thereafter for all subsequent (delta) updates ESH switches to the use of the TREX Memory-Delta index. This you can see in transaction TREXADMIN->Index Admin.
Such memory-delta indices are incompatible with Master–Slave arrangements so such a configuration cannot be utilised in conjunction with ESH (SAP Note 1628169).
Additionally all the ESH indices must reside on the same master – otherwise there are problems in implementing the index joins. Therefore the only recommended configuration for TREX in the ESH scenario is a (single IndexServer) Master only or Master-Backup scaled vertically (please contact the H/W partner for scaling estimations). You can check this here.
These memory-delta indices need to be merged periodically with the main index and this traditionally has been managed by the Memwatch function in TREX. However it has been found that this process can interfere with the indexing process leading to instability in TREX and possible data loss.
Therefore the (only) recommended method of managing the memory-deltas for all applications is that detailed in the KBA below.
KBA 2196916 How to schedule cron job for merging delta indexes
To that end there is a script that you can use to check that TREX conforms to these recommendations; this is attached to the note below. It can create the CRON process in the system if it is not there already and disable the existing (Memwatch) merge.
2227741 TREX 710: check of the TREX settings for the Enterprise/embedded Search scenario
Please download the script and unpack it to the /usr/sap/<sid>/trxnn/exe/python_support directory and from there run it as
Please follow up on any advisements from the script.
What it will check:
- Correct TREX configuration ie. No Master-Slave.
- Preloading enabled for performance.
- Indexing threads.
- RFC Threads.
- Disable Memwatch merge and enable CRON (manual CRONTAB.ini merge scheduling still required).
- Preprocessor unescape setting.
- Turns off IndexUsage Collector.
We would recommend that this is executed in all TREX for ESH systems to ensure a common basic minimum setup.
Finally in older releases of Enterprise Search authorisation indices are not implemented as memory-deltas.
Please implement all the notes detailed in the note below to effect the implementation of authorisation updates as memory-deltas also.
2199115 TREX 710: ESH scenario - indexing of the authorizations indexes
TREX Product Support.