[openstack-dev] [TripleO][CI] Elastic recheck bugs

Ben Nemec openstack at nemebean.com
Mon May 9 20:44:32 UTC 2016


On 05/09/2016 02:32 PM, Clark Boylan wrote:
> On Mon, May 9, 2016, at 10:22 AM, Sagi Shnaidman wrote:
>> Hi, all
>>
>> I'd like to enable elastic recheck on TripleO CI and have submitted
>> patches
>> for refreshing the tracked logs [1] (please review) and for timeout case
>> [2].
>> But according to Derek's comment behind the timeout issue could be
>> multiple
>> issues and bugs, so I'd like to clarify - what are criteria for elastic
>> recheck bugs?
>>
>> I thought about those markers:
>>
>> Nova:
>> 1) "No valid host was found. There are not enough hosts"
>> Network issues:
>> 2) "Failed to connect to trunk.rdoproject.org" OR "fatal: The remote end
>> hung up unexpectedly"  OR "Could not resolve host:"
>> Ironic:
>> 3) "Error contacting Ironic server:"
>> 4) "Introspection completed with errors:"
>> 5) ": Introspection timeout"
>> 6) "Timed out waiting for node "
>> Glance:
>> 7) "500 Internal Server Error: Failed to upload image"
>> crm_resource:
>> 8) "crm_resource for openstack "
>>
>> and various puppet errors.
>>
>> However almost all of these messages could have different root causes,
>> except of network failures. Easy to fix bug doesn't make to submit there,
>> because they will be fixed yet before recheck patch will be merged.
>> So, could you please think about right criteria of bugs for elastic
>> recheck?
> 
> I have reviewed the change to index these logs, you need to create log
> files that are indexable before you can index them. Timestamps are a
> huge thing here as we use delayed processing of the log files.
> 
> As far as criteria for elastic recheck bugs you are correct, the
> preference is to identify specific failures in the service logs. For
> example 500 internal server error: Failed to upload image is what the
> client sees but the server should specifically log what the error was in
> the service logs. The elastic recheck bug should be based on those
> service logs.

Sounds like this is a good time to finally start extracting our logs out
of the tarballs we currently generate:
https://review.openstack.org/#/c/314272




More information about the OpenStack-dev mailing list