[jboss-jira] [JBoss JIRA] (JGRP-1799) RpcDispatcher test fails when working with large values

[Richard Achmatowicz (JIRA)]

    [ https://issues.jboss.org/browse/JGRP-1799?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12951102#comment-12951102 ] 

Richard Achmatowicz edited comment on JGRP-1799 at 3/7/14 7:53 AM:
-------------------------------------------------------------------

Bela, there are still failures of this test at the highest size which take more than 30 seconds on slow machines.
Can we increase this timeout to 60 seconds, from 30 seconds, as it is another case of a test which should eventually return a value and not within any specific time bound.

For example, on Windows 2008:
{noformat}
------------- testLargeReturnValue -----------

testing with 10000 bytes
test took: 34 ms
rsps:
10000 bytes from A
10000 bytes from B
10000 bytes from C

testing with 20000 bytes
test took: 33 ms
rsps:
20000 bytes from A
20000 bytes from B
20000 bytes from C

testing with 40000 bytes
test took: 35 ms
rsps:
40000 bytes from A
40000 bytes from B
40000 bytes from C

testing with 80000 bytes
test took: 880 ms
rsps:
80000 bytes from A
80000 bytes from B
80000 bytes from C

testing with 100000 bytes
test took: 3574 ms
rsps:
100000 bytes from A
100000 bytes from B
100000 bytes from C

testing with 200000 bytes
test took: 5448 ms
rsps:
200000 bytes from A
200000 bytes from B
200000 bytes from C

testing with 400000 bytes
test took: 9975 ms
rsps:
400000 bytes from A
400000 bytes from B
400000 bytes from C

testing with 800000 bytes
test took: 18045 ms
rsps:
800000 bytes from A
800000 bytes from B
800000 bytes from C

testing with 1000000 bytes
test took: 22037 ms
rsps:
1000000 bytes from A
1000000 bytes from B
1000000 bytes from C

testing with 2000000 bytes
test took: 30074 ms
rsps:
{noformat}

There may also be GC activity in play here.
                
      was (Author: rachmato):
    Bela, there are still failures of this test at the highest size which take more than 30 seconds on slow machines.
Can we increase this timeout to 60 seconds, from 30 seconds, as it is another case of a test which should eventually return a value and not within any specific time bound.

For example, on Windows 2008:
{noformat}
------------- testLargeReturnValue -----------

testing with 10000 bytes
test took: 34 ms
rsps:
10000 bytes from A
10000 bytes from B
10000 bytes from C

testing with 20000 bytes
test took: 33 ms
rsps:
20000 bytes from A
20000 bytes from B
20000 bytes from C

testing with 40000 bytes
test took: 35 ms
rsps:
40000 bytes from A
40000 bytes from B
40000 bytes from C

testing with 80000 bytes
test took: 880 ms
rsps:
80000 bytes from A
80000 bytes from B
80000 bytes from C

testing with 100000 bytes
test took: 3574 ms
rsps:
100000 bytes from A
100000 bytes from B
100000 bytes from C

testing with 200000 bytes
test took: 5448 ms
rsps:
200000 bytes from A
200000 bytes from B
200000 bytes from C

testing with 400000 bytes
test took: 9975 ms
rsps:
400000 bytes from A
400000 bytes from B
400000 bytes from C

testing with 800000 bytes
test took: 18045 ms
rsps:
800000 bytes from A
800000 bytes from B
800000 bytes from C

testing with 1000000 bytes
test took: 22037 ms
rsps:
1000000 bytes from A
1000000 bytes from B
1000000 bytes from C

testing with 2000000 bytes
test took: 30074 ms
rsps:
{noformat}
                  
> RpcDispatcher test fails when working with large values
> -------------------------------------------------------
>
>                 Key: JGRP-1799
>                 URL: https://issues.jboss.org/browse/JGRP-1799
>             Project: JGroups
>          Issue Type: Bug
>    Affects Versions: 3.2.13
>         Environment: RHEL, Win, Solaris
>            Reporter: Richard Achmatowicz
>            Assignee: Bela Ban
>             Fix For: 3.2.13
>
>
> The two tests:
> * testLargeReturnValue
> * testLargeReturnValueUnicastCall
> make RPC calls with values which are increasingly large. 
> The values used are in this range:
> {noformat}
> SIZES={10000, 20000, 40000, 80000, 100000, 200000, 400000, 800000,1000000, 2000000, 5000000}
> {noformat}
> The tests have been see to fail with the values 1000000, 2000000 and 5000000, always with the same error in each case. 
> In the case of testLargeReturnValue, the test fails because one of the returned values from the RPC is null.
> In the case of testLargeReturnValueUnicastCall, the test fails due to a timeout while sending the RPC.
>  

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira