incubator-cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Abhinandan Prateek" <aprat...@apache.org>
Subject Re: Review Request: CLOUDSTACK-803: HA gets triggered even when the host investigator is unable to determine the state of the host
Date Fri, 18 Jan 2013 11:47:43 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/8853/#review15486
-----------------------------------------------------------

Ship it!


Ship It!

- Abhinandan Prateek


On Jan. 7, 2013, 8:26 a.m., Koushik Das wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/8853/
> -----------------------------------------------------------
> 
> (Updated Jan. 7, 2013, 8:26 a.m.)
> 
> 
> Review request for cloudstack, Abhinandan Prateek and Alex Huang.
> 
> 
> Description
> -------
> 
> CS tries to do HA on VMs if it finds that the host on which they are running is down.
HA should only happen when CS is able to determine (host investigators return status as 'Down')
that host is down. Currently due to a bug in code even if the hosts investigators are unable
to determine state (returns 'null'), HA gets triggered.
>  
> This results in unnecessary HA of VMs.
> 
> The fix is to not do HA in case the state of host cannot be determined.
> 
> 
> This addresses bug CLOUDSTACK-803.
> 
> 
> Diffs
> -----
> 
>   server/src/com/cloud/agent/manager/AgentManagerImpl.java 8141fdb 
> 
> Diff: https://reviews.apache.org/r/8853/diff/
> 
> 
> Testing
> -------
> 
> I did the following manual testing
> 
> -	Setup advanced zone with XS host
> -	Created a VM
> -	Pulled out network cable for MS
> -	Checked in the debugger that the DisconnectTask is running
> -	This calls  handleDisconnectWithInvestigation in AgentManagerImpl
> -	In the debugger saw that the following line of code is returning state as null 'final
Status determinedState = investigate(attache);'
> -	So the warning message added is seen 's_logger.warn("Agent state cannot be determined,
do nothing");'
> -	The above steps continues in a loop till the time I plug the network cable back after
which again the status of the host gets correctly determined by the investigator.
> 
> If either the XS host or MS is disconnected from network then investigator cannot determine
the state of the host and so based on the fix HA won’t be triggered.
> 
> Sowmya from Citrix QA team also did additional testing for this on a private branch.
> 
> 
> Thanks,
> 
> Koushik Das
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message