Hi Jesse,

thanks for your reply.

> > jenkins execs the readlink program every time a symlink is followed, which 
> > is probably the cause for the massive delays
> 
> You are speculating, or you measured this (minimally with a series of thread 
> dumps)?

The readlink invocation is "measured" (observed) - this being the reason
for the delay is speculation, though.

> > upgrading to Java7, which can follow readlinks via the nio.* classes
> 
> This is what I would recommend. Not sure what IBM’s support policy is, but 
> Oracle has dropped unpaid support for JDK 6.
> 
> > I'm not sure, however, whether Jenkins would use the nio.* methods instead 
> > of readlink here

In the meantime I've tried 1.509.1-LTS on AIX on java7. Startup and
(initial) list view loading times were reduced by a factor of around 2
(which is already enough in my case for not seeing timeouts).

> It should; if it does not, please investigate ASAP.
> 
> > Patch jna-posix to support AIX
> 
> There is already PR #770 to update to jnr-posix, which at least has 
> AIX-specific classes in it; please help evaluate this PR to see if it 
> improves performance on AIX.

Thanks for the hint - from looking at the Jenkins code, it looks like
Jenkins tries the java7 methods first, then GNU libc, then jna-posix.

Best regards,

Martin

-- 
You received this message because you are subscribed to the Google Groups 
"Jenkins Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.


Reply via email to