Any reason the messages can't be queued? In my mind, the benefits would outweigh the unbounded nature of it. Erik -----Original Message----- From: infinispan-dev-bounces(a)lists.jboss.org [mailto:infinispan-dev-bounces@lists.jboss.org] On Behalf Of Pedro Ruivo Sent: Wednesday, March 13, 2013 9:29 AM To: ispn-dev Subject: [infinispan-dev] DefaultExecutorFactory and rejection policy Hi I'm working on ISPN-2808 (https://issues.jboss.org/browse/ISPN-2808) and I noticed that the DefaultExecutorFactory is creating the executor service with an Abortpolicy. Is there any particular reason for that? In the new thread pool needed by ISPN-2808, I cannot have the messages (i.e. the runnables) discarded, because it can cause some inconsistent state and even block all the cluster. I have set in my branch a CallerRunPolicy. If you see any issue with this let me know. Cheers, Pedro _______________________________________________ infinispan-dev mailing list infinispan-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/infinispan-dev

By default, the thread pools are bounded and we have nothing to prevent the it to be full. currently, the regular message are not processed by the thread pool. this can be solved when a smarter dispatcher/ordering mechanism is implemented. On 03/13/2013 02:07 PM, Bela Ban wrote:

Blocking OOB threads is the thing we want to avoid, remember? I am used do synchronous replication, there we could hold a semaphore on requests to each node. Going down before each RPC and up after that would be pretty simple and having this mechanism in RPC layer it would block only request, not response messages. Are there any shoot & forget messages sent from Infinispan which could not be applied in this way? Regarding the async replication, I am not sure whether the request is just processed in non-caller thrad or if there is anything more. Radim ----- Original Message ----- | From: "Adrian Nistor" <anistor(a)redhat.com&gt; | To: "infinispan -Dev List" <infinispan-dev(a)lists.jboss.org&gt; | Sent: Wednesday, March 13, 2013 3:37:23 PM | Subject: Re: [infinispan-dev] DefaultExecutorFactory and rejection policy | | What about creating a RejectedExecutionHandler that blocks the | calling | thread if the thread pool has reached the max (and all threads are | busy) | and the associated queue is also full? See | org.hibernate.search.batchindexing.impl.Executors.BlockPolicy | | Some time ago I was thinking we need this for infinispan's thread | pools | too because of some very insidious test suite failures caused by | dropped | tasks and created ISPN-2438 but never got to experiment with this. | The | solution used for the moment was to increase the pool a bit for unit | tests... | | Cheers | | On 03/13/2013 04:18 PM, Pedro Ruivo wrote: | > By default, the thread pools are bounded and we have nothing to | > prevent | > the it to be full. | > | > currently, the regular message are not processed by the thread | > pool. | > this can be solved when a smarter dispatcher/ordering mechanism is | > implemented. | > | > On 03/13/2013 02:07 PM, Bela Ban wrote: | >> You mean an unbounded queue ? This would be bad if the insertion | >> rate of | >> messages is higher than the processing rate. | >> | >> The problem with the Infinispan pool is that we cannot simply | >> discard | >> messages; they won't get retransmitted, as they're been delivered | >> by | >> JGroups. | >> | >> If we end up with our own ordering of messages in the ISPN pool, | >> then a | >> rejection policy of "run" would also cause harm, as it would | >> destroy the | >> ordering (except for OOB messages)... | >> | >> On 3/13/13 2:52 PM, Erik Salter wrote: | >>> Any reason the messages can't be queued? In my mind, the | >>> benefits would | >>> outweigh the unbounded nature of it. | >>> | >>> Erik | >>> | >>> -----Original Message----- | >>> From: infinispan-dev-bounces(a)lists.jboss.org | >>> [mailto:infinispan-dev-bounces@lists.jboss.org] On Behalf Of | >>> Pedro Ruivo | >>> Sent: Wednesday, March 13, 2013 9:29 AM | >>> To: ispn-dev | >>> Subject: [infinispan-dev] DefaultExecutorFactory and rejection | >>> policy | >>> | >>> Hi | >>> | >>> I'm working on ISPN-2808 | >>> (https://issues.jboss.org/browse/ISPN-2808) and I | >>> noticed that the DefaultExecutorFactory is creating the executor | >>> service | >>> with an Abortpolicy. | >>> | >>> Is there any particular reason for that? | >>> | >>> In the new thread pool needed by ISPN-2808, I cannot have the | >>> messages (i.e. | >>> the runnables) discarded, because it can cause some inconsistent | >>> state and | >>> even block all the cluster. | >>> | >>> I have set in my branch a CallerRunPolicy. If you see any issue | >>> with this | >>> let me know. | >>> | >>> Cheers, | >>> Pedro | >> | > _______________________________________________ | > infinispan-dev mailing list | > infinispan-dev(a)lists.jboss.org | > https://lists.jboss.org/mailman/listinfo/infinispan-dev | | _______________________________________________ | infinispan-dev mailing list | infinispan-dev(a)lists.jboss.org | https://lists.jboss.org/mailman/listinfo/infinispan-dev |

----- Original Message ----- | From: "Dan Berindei" <dan.berindei(a)gmail.com&gt; | To: "infinispan -Dev List" <infinispan-dev(a)lists.jboss.org&gt; | Sent: Thursday, March 14, 2013 10:03:10 AM | Subject: Re: [infinispan-dev] DefaultExecutorFactory and rejection policy | | On Thu, Mar 14, 2013 at 9:32 AM, Radim Vansa < rvansa(a)redhat.com > | wrote: | | | | Blocking OOB threads is the thing we want to avoid, remember? | | Well, you have to block somewhere... | | I like Adrian's solution, because it's a lot better than | CallerRunsPolicy: it's blocking the OOB thread until any other | command finishes executing, not until one particular command | finishes executing. I don't like caller-runs policy either. OOB threads shouldn't be waiting for anything and executing a command within OOB thread could cause that. In our problem, it would only increase the OOB threadpool size by the ispn thread pool size and cause some overhead to it. We should always have some OOB threads able to process the responses. | | | ... | | I don't think you can throttle on the sender, because you don't know | how many threads the recipient should allocate per sender in the | Infinispan thread pool. You don't need to know how many threads exactly are executing on the receiver, because you would have an unbounded queue there which will, sooner or later, process the messages. The semaphore is there because of throttling so that you never overwhelm the recipient with requests. I don't insist on semaphore-like synchronization, but the receiver should provide some feedback that it's not able to process more messages and the sender should be responsible for abiding it. AND, the feedback should be provided in a way that the node is still able to process other messages. If it is a jammed signal message broadcast after the queue length grows beyond some limit, and then all messages on senders should be postponed, or divine intervention, it's just a performance issue. But if there a situation where the node is not able to process any more messages (on JGroups level), the OOB issue won't be solved, because there may be a reply other message is waiting for that is never processed. | | E.g. in non-tx concurrent mode, for each user-initiated request, the | primary owner sends out numOwners requests to the other nodes. If | the cache is also replicated, you have the coordinator sending | clusterSize requests to the other nodes for each user-initiated | request. (So for each request that a node receives from the | coordinator, it will receive exactly 0 requests from the other | nodes.) | | | If you start considering that each cache should have its own | semaphore, because they need different numbers of threads, then | finding an appropriate initial value is pretty much impossible. | Sorry, I miss the point here. It's not necessary to be exactly able to tell how many messages may go from one node to another, you can just use a common sense to limit the maximum amount of request that should be processed concurrently between the two nodes until we say "Hey, give the poor node a break, it's not responding because it's busy and another message would not help right now". Radim