What is the max size of heap size that can configured in repotools-jts. sh to rebuilt indices
Hi,
We are trying to perform RTC upgrade from 5.0.2 to 6.0.6.1. With RTC 5.0.2 I have currently 100gb (JTS+CCM) indices. I need to rebuild these indices post upgrade. As I have huge amount of indices I would need to increase the heap memory limit in repotools-jts.sh and repotools-ccm.sh. I want to understand what is the maximum size I can configure. I am trying to configure 20G as follows
export REPOTOOLS_MX_SIZE=20000M
Is this correct? I was told that it may take days or week to built such huge indices. Is that true? How can I speed us the rebuilding indices process.
Thanks!
2 answers
Hi Aastha,
- You can improve the network latency from your application server to the database.
- If you have a larger indices large you can speed up the repotools-reindex command by:
a. Stopping server.
b. In [JAZZ_INSTALL]/server/repotools-jts or /repotools-rm file locate the following properties:
VMARGS="$VMARGS -Xmx1500M"
#VMARGS="$VMARGS -Xgcpolicy:gencon"
Modify (depending on the size of your data and memory available on the server) as follows, for example:
VMARGS="$VMARGS -Xmx4096M"
VMARGS="$VMARGS -Xms4096M"
VMARGS="$VMARGS -Xmn1024M"
VMARGS="$VMARGS -Xgcpolicy:gencon"
- If you have a larger indices large you can speed up the repotools-reindex command by:
a. Stopping server.
b. In [JAZZ_INSTALL]/server/repotools-jts or /repotools-rm file locate the following properties:
VMARGS="$VMARGS -Xmx1500M"
#VMARGS="$VMARGS -Xgcpolicy:gencon"
Modify (depending on the size of your data and memory available on the server) as follows, for example:
VMARGS="$VMARGS -Xmx4096M"
VMARGS="$VMARGS -Xms4096M"
VMARGS="$VMARGS -Xmn1024M"
VMARGS="$VMARGS -Xgcpolicy:gencon"
When we upgraded repos to 6.0.6.1 we simply put the index folder in a separate location which forced the application to rebuild indexes in the background.
6.0.6.1 New and Noteworthy has this to say:
Full-text search indexes must be rebuilt when you upgrade to 6.0.6.1 or later
In the 6.0.6.1 release, the Apache Lucene full-text search engine was upgraded to a more recent version. This upgrade requires all existing fulltext search indexes that were created in 6.0.6 or earlier to be rebuilt. Full-text search indexes must be rebuilt for both the CLM servers and the Rational Team Concert Client for Eclipse IDE.
Typically, when you upgrade an installation by using the upgrade scripts, the full-text search indexes from the earlier installation are copied to the new installation. However, the upgrade scripts in 6.0.6.1 and later do not copy previous indexes from version 6.0.6 or earlier (including interim fixes). Instead, the new server automatically rebuilds the indexes in the background after you start it. Results of full-text searches might be incomplete until the reindexing operation is finished; the progress of the reindex is periodically written to the server log file.
Alternatively, instead of allowing the server to rebuild full-text search indexes online, you can rebuild them manually by running the repotools-<app> -rebuildTextIndices command after the upgrade is complete, but before you start the server, where <app> is one of am, ccm, or qm. Note that reindexing might be a lengthy process and can take many hours for a large database. In the web client, a system alert in
the banner states that reindexing is in progress.