Introduce Infinispan as cache and hibernate back-end #577

carlosmunoz · 2014-09-03T23:40:47Z

Make infinispan the backend used for all caches (hibernate included) as well as hibernate search indexes. All configuration for caching now falls on standalone.xml. Additionally, these caches can be made to persist to the database meaning it's now possible to have caches that persist server shutdowns (statistics come to mind here), as well as keep them solely in memory (to test hibernate search without needing to create index files)

carlosmunoz · 2014-09-03T23:41:45Z

@seanf Sean, do you mind taking an early look? No configuration details in the PR, I'll create a wiki page for those.

carlosmunoz · 2014-09-05T02:04:26Z

For configuration details (still a very early draft) see https://github.com/zanata/zanata-server/wiki/Infinispan

seanf · 2014-09-12T01:20:56Z

@carlosmunoz Shouldn't we keep the integration tests' configuration as close to production as we can, in terms of persistence?

carlosmunoz · 2014-09-12T01:36:44Z

Might not be a bad idea for the arquillian tests. I'm sure the JdbcStore works with H2. Until we get some serious performance tests I don't think we'll be able to see anything go to the database though, but at least it will validate our configuration.

seanf · 2014-09-12T01:43:48Z

zanata-war/src/main/java/org/zanata/cache/InfinispanCacheWrapper.java

+
+    @Override
+    public V getWithLoader(K key) {
+        // NB: Need to manually implement the cache loader feature


Should this method be synchronised or something?

seanf · 2014-09-12T01:44:40Z

What are the effects of switching from CacheConcurrencyStrategy.READ_WRITE to CacheConcurrencyStrategy.TRANSACTIONAL?

seanf · 2014-09-12T01:49:11Z

zanata-war/src/main/java/org/zanata/service/impl/TranslationStateCacheImpl.java

-        cacheManager.shutdown();
+        // NB Since infinispan is container managed, there's no need to stop the
+        // cache manager with
+        // cacheContainer.stop();


Might as well get rid of the whole method then. One less thing to migrate from Seam to CDI.

carlosmunoz · 2014-09-12T01:54:15Z

What are the effects of switching from CacheConcurrencyStrategy.READ_WRITE to CacheConcurrencyStrategy.TRANSACTIONAL?

From what I've investigated, READ_WRITE is the least performant of all the basic options just because it locks entries and compares versions (hash?) of objects when saving. TRANSACTIONAL is the one designed specifically for distributed caches. It does not lock, but rather relies on the cache to provide consistency. It can be registered with the local JTA transaction and comes with the price of XA. So, it will be less efficient, but if we can use the option to save in the background it might not be noticeable from a response time perspective... might even be faster than our current solution.

seanf · 2014-09-12T01:56:02Z

zanata-war/src/test/java/org/zanata/rest/service/ResourceServiceRestTest.java

    @Override
    protected void prepareResources() {
        MockitoAnnotations.initMocks(this);
-
+        cacheContainer.start();


Shouldn't we call cacheContainer.stop() somewhere too?

seanf · 2014-09-12T01:57:44Z

I won't list all of them, but I think all the tests which start() CacheContainer should also stop() it in an @AfterMethod or something.

seanf · 2014-09-12T01:59:36Z

zanata-war/src/test/resources/arquillian/persistence.xml

+        value="infinispan" />
+      <!-- to use an existing cache manager (Good for external configuration) -->
+      <property name="hibernate.search.infinispan.cachemanager_jndiname"
+        value="java:jboss/infinispan/container/hibernate-search"/>


What about this bit (from the other persistence.xml)?

<property name="hibernate.cache.region.factory_class" value="org.hibernate.cache.infinispan.JndiInfinispanRegionFactory" /> <property name="hibernate.cache.infinispan.cachemanager" value="java:jboss/infinispan/container/hibernate" />

Conflicts: zanata-war/src/test/java/org/zanata/service/impl/StatisticsServiceImplTest.java

carlosmunoz · 2015-01-29T05:17:00Z

@seanf Can we consider merging this into master now that we've branched for release?

carlosmunoz · 2015-02-26T03:49:34Z

@seanf bump

seanf · 2015-02-26T04:11:36Z

@carlosmunoz Looks good, but we might need QA to look again, after the branch is rebased (or master is merged in).

Conflicts: pom.xml zanata-war/pom.xml zanata-war/src/main/java/org/zanata/service/impl/TranslationStateCacheImpl.java

carlosmunoz · 2015-03-11T04:36:38Z

@seanf Care to review the last commits for this pull request?

seanf · 2015-03-11T04:41:55Z

docs/release-notes.md

+<h5>Infrastructure Changes</h5>
+
+Zanata now uses Infinispan as its cache provider, and the cache needs to be configured in Jboss' `standalone.xml` file. Please see the [Infinispan](configuration/infinispan) section for more information.
+


I don't think this pull request updates all SIX! copies of standalone*.xml in the source tree.

Added the configuration to all our standalone.xml files.

seanf · 2015-03-11T04:43:04Z

Otherwise looks fine.

carlosmunoz · 2015-03-11T05:05:17Z

No. We left hibernate search out of infinispan for the time being due to less than desirable performance.

seanf · 2015-03-11T05:24:38Z

👍 Once the tests pass. I would suggest running all functional tests before merging though.

Introduce Infinispan as cache and hibernate back-end Full functional tests passed. Not attempting a squash merge.

carlosmunoz added the Work in progress label Sep 3, 2014

carlosmunoz force-pushed the infinispan branch from 9d4736e to f2ca245 Compare September 3, 2014 23:47

carlosmunoz force-pushed the infinispan branch 4 times, most recently from 8c1a38e to 3a3d8ee Compare September 12, 2014 01:12

seanf reviewed Sep 12, 2014
View reviewed changes

carlosmunoz force-pushed the infinispan branch 7 times, most recently from dcd7aa2 to 671603b Compare September 15, 2014 05:30

carlosmunoz added 2 commits November 28, 2014 15:11

Merge branch 'master' into infinispan

b2c6d3b

Conflicts: zanata-war/src/test/java/org/zanata/service/impl/StatisticsServiceImplTest.java

Refactor initialization logic for cache container.

0501b44

carlosmunoz force-pushed the infinispan branch from 647ac21 to 0501b44 Compare November 28, 2014 05:40

Compilation error fixes.

8d9862f

djansen-redhat removed the Verified label Jan 21, 2015

seanf added the Reviewed label Feb 26, 2015

Merge remote-tracking branch 'origin/master' into infinispan

8dc9f6c

Conflicts: pom.xml zanata-war/pom.xml zanata-war/src/main/java/org/zanata/service/impl/TranslationStateCacheImpl.java

carlosmunoz force-pushed the infinispan branch from 2807aef to 8dc9f6c Compare February 27, 2015 04:06

Carlos A. Munoz added 3 commits March 10, 2015 14:31

Merge remote-tracking branch 'origin/master' into infinispan

17bcbce

Fix maven validation problems.

f6bd5b2

Correct checkstyle errors

e02cd00

carlosmunoz force-pushed the infinispan branch from dd3deeb to e02cd00 Compare March 11, 2015 04:01

seanf reviewed Mar 11, 2015
View reviewed changes

Activate Infinispan configuration documentation.

38cab87

carlosmunoz force-pushed the infinispan branch from ca64ea9 to 38cab87 Compare March 11, 2015 04:42

Add Infinispan configuration to all standalone.xml instances.

72a6b50

carlosmunoz force-pushed the infinispan branch from 55b4e3b to 72a6b50 Compare March 11, 2015 05:43

carlosmunoz added a commit that referenced this pull request Mar 12, 2015

Merge pull request #577 from zanata/infinispan

5decdf0

Introduce Infinispan as cache and hibernate back-end Full functional tests passed. Not attempting a squash merge.

carlosmunoz merged commit 5decdf0 into integration/master Mar 12, 2015

carlosmunoz deleted the infinispan branch March 12, 2015 04:33

seanf mentioned this pull request Mar 13, 2015

Upgrade RichFaces to 4.5.2.Final #646

Closed

carlosmunoz restored the infinispan branch March 9, 2016 06:38

seanf deleted the infinispan branch September 21, 2016 02:22

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Introduce Infinispan as cache and hibernate back-end #577

Introduce Infinispan as cache and hibernate back-end #577

carlosmunoz commented Sep 3, 2014

carlosmunoz commented Sep 3, 2014

carlosmunoz commented Sep 5, 2014

seanf commented Sep 12, 2014

carlosmunoz commented Sep 12, 2014

seanf Sep 12, 2014

seanf commented Sep 12, 2014

seanf Sep 12, 2014

carlosmunoz commented Sep 12, 2014

seanf Sep 12, 2014

seanf commented Sep 12, 2014

seanf Sep 12, 2014

carlosmunoz commented Jan 29, 2015

carlosmunoz commented Feb 26, 2015

seanf commented Feb 26, 2015

carlosmunoz commented Mar 11, 2015

seanf Mar 11, 2015

carlosmunoz Mar 11, 2015

seanf commented Mar 11, 2015

carlosmunoz commented Mar 11, 2015

seanf commented Mar 11, 2015

		<h5>Infrastructure Changes</h5>

		Zanata now uses Infinispan as its cache provider, and the cache needs to be configured in Jboss' `standalone.xml` file. Please see the [Infinispan](configuration/infinispan) section for more information.

Introduce Infinispan as cache and hibernate back-end #577

Introduce Infinispan as cache and hibernate back-end #577

Conversation

carlosmunoz commented Sep 3, 2014

carlosmunoz commented Sep 3, 2014

carlosmunoz commented Sep 5, 2014

seanf commented Sep 12, 2014

carlosmunoz commented Sep 12, 2014

seanf Sep 12, 2014

Choose a reason for hiding this comment

seanf commented Sep 12, 2014

seanf Sep 12, 2014

Choose a reason for hiding this comment

carlosmunoz commented Sep 12, 2014

seanf Sep 12, 2014

Choose a reason for hiding this comment

seanf commented Sep 12, 2014

seanf Sep 12, 2014

Choose a reason for hiding this comment

carlosmunoz commented Jan 29, 2015

carlosmunoz commented Feb 26, 2015

seanf commented Feb 26, 2015

carlosmunoz commented Mar 11, 2015

seanf Mar 11, 2015

Choose a reason for hiding this comment

carlosmunoz Mar 11, 2015

Choose a reason for hiding this comment

seanf commented Mar 11, 2015

carlosmunoz commented Mar 11, 2015

seanf commented Mar 11, 2015