[jira] [Commented] (IGNITE-1094) Ignite.createCache(CacheConfiguration) hangs if some exception occurs during cache initialization

["Vladimir Ozerov (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IGNITE-1094?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16588722#comment-16588722 ] 

Vladimir Ozerov commented on IGNITE-1094:
-----------------------------------------

[~agura], [~Jokser], [~agoncharuk], [~slava.koptilin], [~dpavlov],

Igniters,

It seems that real problem is not fixed. This is true, that after the fix cache with invalid configuration no longer hangs a client. However, it breaks exchange worker on other nodes, and moves it into unrecoverable state. Any subsequent exchange-related operation, such as node stop or cache start/stop will kill the node. Essentially, we just delayed exchange worker death/hang a bit. 

Reproducer - \{{SqlIllegalSchemaSelfTest}}. Technically, this test pass. But note the following:
 # See logs of \{{*Dynamic}} tests - instead of normal stop, node gets killed by failure detector:

{code:java}
[2018-08-22 14:20:39,777][ERROR][exchange-worker-#77%query.SqlIllegalSchemaSelfTest%][IgniteTestResources] Critical system error detected. Will be handled accordingly to configured handler [hnd=class o.a.i.failure.NoOpFailureHandler, failureCtx=FailureContext [type=SYSTEM_WORKER_TERMINATION, err=java.lang.AssertionError: stopping=true, groupName=null, caches=[]]]
java.lang.AssertionError: stopping=true, groupName=null, caches=[]
at org.apache.ignite.internal.processors.cache.CacheGroupContext.singleCacheContext(CacheGroupContext.java:371){code}

 # Similar behavior is observed if one tries to start caches with invalid configuration twice. Pick any \{{*Dynamic}} test and just copy/paste \{{GridTestUtils.assertThrows}} logic one after another. Expected - two expected exceptions, actual - node is killed.

Reopening the ticket. Please feel free to create another ticket to handle it, if you find it more convenient. But my opinion is that we should fix it here, because cluster is broken and it is just a matter of chance that we didn't spot it in original test.

> Ignite.createCache(CacheConfiguration) hangs if some exception occurs during cache initialization
> -------------------------------------------------------------------------------------------------
>
>                 Key: IGNITE-1094
>                 URL: https://issues.apache.org/jira/browse/IGNITE-1094
>             Project: Ignite
>          Issue Type: Bug
>          Components: cache
>            Reporter: Sergey Evdokimov
>            Assignee: Vyacheslav Koptilin
>            Priority: Major
>              Labels: Muted_test
>             Fix For: 2.7
>
>
> User can pass broken configuration, for example, store factory that throws exception from create() method. I created test to demonstrate the problem. See IgniteDynamicCacheStartSelfTest#testBrokenStoreFactory in 'ignite-1094' branch 



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)