Cache nodes in cluster write to shared DB store at the same time in batching mode even "shared" is set to ture

[zband]

Hi,

I have replicated async cache configured in a cluster with shared db store, I have set the db store attribute shared="true" so only the node where the modification originated will write to the db. However after I enabled batching mode, I noticed all the nodes started write to db at the same time and getting unique constraint error from some nodes, even when I just add a cache entry without using a transaction. Below is the config for my cache, not sure if I did anything wrong. Can you please advise?

<replicated-cache name="repl" mode="ASYNC">
      <transaction mode="BATCH"/>
     <locking acquire-timeout="30000" concurrency-level="1000" striping="false"/>
        <eviction strategy="LRU" max-entries="10000"/>
        <persistence passivation="false">
    <string-keyed-jdbc-store shared="true">
       <data-source jndi-url="/jdbc/cacheregion" />
       <string-keyed-table drop-on-exit="true" create-on-start="true" prefix="${jdbc.table.name.prefix}">
          <id-column name="ID" type="VARCHAR(255)" />
          <data-column name="DATA" type="BLOB" />
          <timestamp-column name="TIMESTAMP" type="NUMBER(19)" />
       </string-keyed-table>
       <write-behind flush-lock-timeout="1" modification-queue-size="1024" shutdown-timeout="25000" thread-pool-size="4"/>
    </string-keyed-jdbc-store>
  </persistence>
     </replicated-cache>

Thanks

Tao

[ryanemerson]

Hi Tao,

I was able to reproduce the shared flag being ignored in 8.2.3.Final and It seems that there is an underlying issue here.  What version of infinispan are you currently utilising?

I have created [1] to track the issue and I will look into this soon.

[1] [ISPN-6907] Replicated cache with BATCH Transactions do not follow shared store semantics - JBoss Issue Tracker

Thanks

Ryan

[zband]

Hi Ryan,

Thanks for your prompt reply and the confirmation on the bug, this has been troubling us for few days. I am currently using 7.1.0, I have tested multiple versions including the latest, they seem all have the same issue. Is there any quick fix or ways to walk around this?

Thanks

Tao