Thanks for looking into this, Josh - really appreciate it.

The NewerTableAlreadyExistsException occurs if a test derived from
BaseClientManagedTimeIT does not set the CURRENT_SCN_ATTRIB property on the
connection when creating a table. This will end up occurring in a
*different* test than the test that forgets to do this, as it happens when
the same named table is created. I fixed the one I found in CreateTableIT
in this[1] commit, but there may be more.

We have a few options to make this less brittle and easier to track down:
- in BaseTest.deletePriorTables(), we can do an assert outside of the loop
that confirms that there are no more user tables remaining. At least then
we'd know which test was doing this.
- convert tests to be derived from BaseHBaseManagedTimeTableReuseIT and use
generated table names as was done in this[2] pull request. This is the best
long term solution as it'll make our tests more reliable and make them 10x
faster. There are a lot of tests to convert, but perhaps we can divide them
up, do them little by little?

Thanks,
James

[1]
https://git-wip-us.apache.org/repos/asf?p=phoenix.git;a=commit;h=0fbce9cd3122a16a890926e0b0042931ae8e03c4
[2] https://github.com/apache/phoenix/pull/158


On Thu, Apr 28, 2016 at 4:11 PM, Josh Elser <[email protected]> wrote:

> Also, didn't mean to ignore you Sergey. I haven't seen anything at the
> moment which pointed to getting more info from -X (general flakiness from
> the ASF jenkins boxes is... the norm).
>
> We can definitely do this if we think there is something concrete that
> we're missing.
>
>
> Sergey Soldatov wrote:
>
>> Can we add -X to the mvn? All recent failures happen right before
>> integration tests started without any reasonable error message.
>> Possible an additional debug info would help.
>>
>> Thanks,
>> Sergey
>>
>> On Tue, Apr 26, 2016 at 3:47 PM, James Taylor<[email protected]>
>> wrote:
>>
>>> Anyone know why our Jenkins builds keep failing? Is it environmental and
>>> is
>>> there anything we can do about it?
>>>
>>> Thanks,
>>> James
>>>
>>

Reply via email to