[ 
https://issues.apache.org/jira/browse/IGNITE-8619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16518083#comment-16518083
 ] 

Ivan Fedotov edited comment on IGNITE-8619 at 6/20/18 4:38 PM:
---------------------------------------------------------------

[~dpavlov], [~ein], Hi!
I investigated  this issue and found that locally tests failed only when ssh 
disconnected per timeout. I assume that ssh timeout on TeamCity could be set 
incorrectly: as one of StartStopNodes tests is started, ssh could be 
disconnected per timeout.  
Change ssh timeout on server side is easy, we should add some lines in 
~/.ssh/sshd_config file: 
ClientAliveInternal n - the server send the clients «null packet» every n 
seconds to keep connection alive,
ClientAliveCountMax m - «null packet» will be send m times.
I suppose that it can fix flucky tests on ssh connection. What do you think?



was (Author: ivanan.fed):
[~dpavlov], [~ein], Hi!
I investigated  this issue and found that locally tests failed only when ssh 
disconnected per timeout. I assume that ssh timeout on TeamCity could be set 
incorrectly: as one of StartStopNodes test is started, ssh could be 
disconnected per timeout.  
Change ssh timeout on server side is easy, we should add some lines in 
~/.ssh/sshd_config file: 
ClientAliveInternal n - the server send the clients «null packet» every n 
seconds to keep connection alive,
ClientAliveCountMax m - «null packet» will be send m times.
I suppose that it can fix flucky tests on ssh connection. What do you think?


> 'Remote node could not start' in ssh connection
> -----------------------------------------------
>
>                 Key: IGNITE-8619
>                 URL: https://issues.apache.org/jira/browse/IGNITE-8619
>             Project: Ignite
>          Issue Type: Bug
>            Reporter: Ivan Fedotov
>            Assignee: Ivan Fedotov
>            Priority: Major
>
> Now there is a problem with launch remote node via ssh. Initially was an 
> assumption that it's due to remote process has not enough time to write 
> information into log: 
> [IGNITE-8085|https://issues.apache.org/jira/browse/IGNITE-8085]. But this 
> correction didn't fix [TeamCity 
> |https://ci.ignite.apache.org/project.html?projectId=IgniteTests24Java8&testNameId=6814497542781613621&tab=testDetails].
>  
> So  it's necessary to make launch remote node via ssh always succesful.



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

Reply via email to