[jbosscache-dev] Fixing PojoCache tutorial - exception from getState

[Brian Stansberry]

Re: the handling of getState, if the state transfer is not done via RPC 
but rather via the JGroups state transfer protocols, the sender cache 
has no way to know what the receiver wants (in memory, persistent).  It 
can only rely on its own configuration to decide what to send.


Galder Zamarreno wrote:
> Alternatively, if that code is behaving as it should, why is there that 
> "hack" to prevent in memory state transfer for the GUI? So that it 
> always starts empty?
> 
> Galder Zamarreno wrote:
>> Hi,
>>
>> I'm trying to fix the tutorial for PojoCache. This works slightly 
>> different because you have an instance of the GUI and then the code 
>> entered via ./runShellDemo.sh
>>
>> I have noted that in JBossCacheView, there following happens before 
>> starting the cache:
>>
>> // hack to prevent a state transfer for now
>> cache.getConfiguration().setFetchInMemoryState(false);
>>
>> pojocache.bsh which is loaded via ./runShellDemo.sh still uses the 
>> same replSynch-service.xml descriptor but it does not set fetch in 
>> memory to false.
>>
>> So, if you start the GUI first, and then execute ./runShellDemo.sh and 
>> then type sourceRelative("pojocache.bsh");, you get an exception like 
>> this:
>>
>> "Caused by: org.jboss.cache.CacheException: Cache instance at 
>> 127.0.0.1:33058 cannot integrate state since state provider could not 
>> provide state due to org.jboss.cache.CacheException: Cache instance at 
>> 127.0.0.1:33056 is not configured to provide state"
>>
>> Now, i'm debating the suitability of the following code in 
>> StateTransferManager.getState:
>>
>> if (!fetchPersistentState && !fetchTransientState)
>> {
>> e = new CacheException("Cache instance at " + cache.getLocalAddress() 
>> + " is not configured to provide state");
>> }
>>
>> Documentation says: "FetchInMemoryState: Whether or not to acquire the 
>> initial in-memory state from existing members. "
>>
>> It does not say anything about giving a state. There's nothing saying 
>> that a cache not configured to fetch should not be able to give it. I 
>> mean, a cache could potentially be configured not to retrieve 
>> transient state on startup, but another cache could be configured to 
>> do so and should be able to retrieve it from the first cache started, 
>> shouldn't it?
>>
>> Cheers,
>>
> 


-- 
Brian Stansberry
Lead, AS Clustering
JBoss, a division of Red Hat
brian.stansberry at redhat.com