On Dec 12, 2012, at 1:08 AM, Shawn Heisey <[email protected]> wrote:

> On 12/11/2012 8:19 PM, Mark Miller wrote:
>> I think there are two different issues at least going on - mostly around 
>> replication. Just happens the solrcloud tests do a lot of that, so they tend 
>> to pick up on issues there.
>> 
>> I look and see if they are slower than normal - my test runs are looking a 
>> little longer lately, but not by too much.
>> 
>> The BasicDistributedZk* test are actually many tests rolled up into one - 
>> mostly to save time on jetty construction/destruction steps. That is part of 
>> why they are longer. There are a few things that add up - one is that we 
>> start jetties sequentially due to system properyy issues - we also currently 
>> stop them sequentially, just because no one has gotten to doing that in 
>> parallel. There are a variety of other things as well.
>> 
>> These tests end up catching many none solrcloud bugs though - most of the 
>> solrcloud stuff is built on existing features and stresses them more than 
>> usual.
> 
> If the amount of time that these tests is taking is considered normal 
> (consistently just over 4 minutes for the one test/seed I singled out to 
> fail), then I'll just have to live with it.  Since these tests do catch bugs, 
> they are good things to have.
> 
> I'm just feeling annoyance that the build time is dominated by tests for 
> features I don't use, especially because when I have a test failure, it's 
> almost always those tests.  I'll get over it once the problems are found and 
> they stop failing so often.
> 
> I haven't gotten a single successful test run since this morning, even when I 
> just run "ant clean test" from branch_4x/solr.  I think it's time to go to 
> bed now.
> 
> Thanks,
> Shawn

Well, I guess normal depends on who you are - for my machine it's closer to 2 
minutes. If it's 4 minutes for a lot of people, it may make sense to break it 
into 2 sets.

Right now these tests are catching a problem around replication and shutdown 
like I mentioned. The replication test hits the same thing sometimes, but much 
less often.

If you don't use replication, you are right - this probably has no effect on 
you. But they are not SolrCloud specific fails.

- Mark


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to