You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by Mark Miller <ma...@gmail.com> on 2015/02/14 19:16:35 UTC

Solr Tests On Our Jenkins Cluster Not Stable

Tests fails on jenkins seem kind of out of control for quite a while now if
anyone has an interest.

I'll see what I can badapple or quickly fix over the next couple days.


- Mark

Re: Solr Tests On Our Jenkins Cluster Not Stable

Posted by Shalin Shekhar Mangar <sh...@gmail.com>.
I was the one who upgraded trunk to Jetty9 and the SocketProxy gave me a
lot of grief. Then Tim fixed the SocketProxy to get the test suite passing.
Let me see if I can find why we suddenly have so many more of these
exceptions on Jetty9.

On Sun, Feb 15, 2015 at 12:39 AM, Mark Miller <ma...@gmail.com> wrote:

>
> > On Feb 14, 2015, at 1:46 PM, Shalin Shekhar Mangar <
> shalinmangar@gmail.com> wrote:
> >
> > Particularly frequent are the failures from NoHttpResponseException. Do
> you know why?
>
> This used to happen, but very rarely. It was a more common in the issue
> where I took out the httpclient stale check out (not committed).
>
> Some change by someone not long ago (jetty 9? a change from the backup
> issue? IDK) made it start failing always on Windows and more commonly on
> even linux Jenkins runs (my local checkouts have been fine).
>
> Currently, the only thing we can do is retry on these exceptions.
>
> In these particular tests, it’s probably happening much more easily
> because of the Jetty Proxy stuff.
>
> It means the full update has been sent and we got no response - so we
> don’t know if the request was received or not. It can and will happen,
> probably because of the connection pool stuff, there seems to be no way to
> eliminate it entirely that I have found. But it’s rare enough we normally
> do not have to consider - for the jetty proxy tests though, we probably
> have to look at more retries (i put some in the tests that helped when I
> was removing the httpclient stale check, but it didn’t seem to help
> whatever is happening here now).
>
> - Mark
>
> http://about.me/markrmiller
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: dev-help@lucene.apache.org
>
>


-- 
Regards,
Shalin Shekhar Mangar.

Re: Solr Tests On Our Jenkins Cluster Not Stable

Posted by Mark Miller <ma...@gmail.com>.
> On Feb 14, 2015, at 1:46 PM, Shalin Shekhar Mangar <sh...@gmail.com> wrote:
> 
> Particularly frequent are the failures from NoHttpResponseException. Do you know why?

This used to happen, but very rarely. It was a more common in the issue where I took out the httpclient stale check out (not committed).

Some change by someone not long ago (jetty 9? a change from the backup issue? IDK) made it start failing always on Windows and more commonly on even linux Jenkins runs (my local checkouts have been fine).

Currently, the only thing we can do is retry on these exceptions.

In these particular tests, it’s probably happening much more easily because of the Jetty Proxy stuff.

It means the full update has been sent and we got no response - so we don’t know if the request was received or not. It can and will happen, probably because of the connection pool stuff, there seems to be no way to eliminate it entirely that I have found. But it’s rare enough we normally do not have to consider - for the jetty proxy tests though, we probably have to look at more retries (i put some in the tests that helped when I was removing the httpclient stale check, but it didn’t seem to help whatever is happening here now).

- Mark

http://about.me/markrmiller


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org


Re: Solr Tests On Our Jenkins Cluster Not Stable

Posted by Shalin Shekhar Mangar <sh...@gmail.com>.
Hi Mark,

I'm on board. Let's see if we can fix some of these failures. Particularly
frequent are the failures from NoHttpResponseException. Do you know why?
There are other failures related to TestReplicationHandler and
TestReplicationHandlerBackup which I'll take up.

On Sat, Feb 14, 2015 at 11:46 PM, Mark Miller <ma...@gmail.com> wrote:

> Tests fails on jenkins seem kind of out of control for quite a while now
> if anyone has an interest.
>
> I'll see what I can badapple or quickly fix over the next couple days.
>
>
> - Mark
>



-- 
Regards,
Shalin Shekhar Mangar.