[Gluster-devel] Is there a Regression Test cycle exist?

Geoff Kassel gkassel at users.sourceforge.net
Fri Mar 27 00:00:28 UTC 2009


Hi,
   I'd also like to put my vote in for (more) regression tests - especially 
scripted/automated tests of the 'make test' variety which the public can use.

   Given recent data integrity issues, this would help to instill confidence 
in GlusterFS releases by allowing users to test a new release build before 
putting it into use.

   It would also allows users of more exotic environments (i.e. kernels with 
PaX or SELinux, Darwin, FreeBSD, etc) to determine exactly where their 
environments cause issues. This would make it much easier for the GlusterFS 
developers to maintain compatibility with these platforms.

Kind regards,

Geoff Kassel.

On Fri, 27 Mar 2009, Kirby Zhou wrote:
> Maybe unit test is too difficult to add to existing code.
>
> I think the following classical cases should be considered integrated into
> the whole suite regression test.
>
>
>
> 1.       Test Environments
>
> a)         4 nodes with DHT.
>
> b)         A with all performance xlators.
>
> c)         B with all feature xlators.
>
> d)         4*2 nodes with DHT and AFR.
>
> e)         D with all performance xlators.
>
> f)          E with all feature xlators.
>
> 2.       Test Case
>
> a)         Small file test. Build kernel or gcc or apache over Env.A-F
>
> b)         Big file test. Run iozone with file larger than 2G.
>
> c)         Huge file test. Construct some huge sparse file larger than 2T.
>
> d)         Failover test. Shutdown 3 node of Env.D~F while Case.A~ Case.C
> running.
>
>
>
> Thanks





More information about the Gluster-devel mailing list