Please review: nunc stans 57, job state machine

[William Brown]

https://fedorahosted.org/nunc-stans/ticket/57

https://fedorahosted.org/nunc-stans/attachment/ticket/57/0001-Ticket-57-Impl...

https://fedorahosted.org/nunc-stans/attachment/ticket/57/0002-Ticket-57-Upda...

May not apply without #51, #54, #56

[William Brown]

On Wed, 2016-07-13 at 14:47 +1000, William Brown wrote:
> https://fedorahosted.org/nunc-stans/ticket/57
>
> https://fedorahosted.org/nunc-stans/attachment/ticket/57/0001-Ticket-57-Impl...
>
> https://fedorahosted.org/nunc-stans/attachment/ticket/57/0002-Ticket-57-Upda...
>
> May not apply without #51, #54, #56

Mentioned it in the ticket, but this builds and works with git master of
DS with ldclt. So the code is "reliable" from that aspect.

[Mark Reynolds]

On 07/13/2016 02:28 AM, William Brown wrote:
> On Wed, 2016-07-13 at 14:47 +1000, William Brown wrote:
> > https://fedorahosted.org/nunc-stans/ticket/57
> >
> > https://fedorahosted.org/nunc-stans/attachment/ticket/57/0001-Ticket-57-Impl...
> >
> > https://fedorahosted.org/nunc-stans/attachment/ticket/57/0002-Ticket-57-Upda...
> >
> > May not apply without #51, #54, #56
>
> Mentioned it in the ticket, but this builds and works with git master of
> DS with ldclt. So the code is "reliable" from that aspect.

Can you please run the attached connection stress test script as well
(for 24 hours)?  I'm worried that you are introducing a lot of changes
like might cause regressions.

You need to already have a server instance setup with these entries and
the idle timeout set to 10 seconds.  You will also need to tweak the
rootdn and password in the script.  I know this script needs more
polishing but it stresses the server in ways ldclt can not.

uid=mark,dc=example,dc=com

cn=group1,dc=example,dc=com

...

cn=group10,dc=example,dc=com

Thanks,

Mark
> --
> 389-devel mailing list
> [e-mail]
> https://lists.fedoraproject.org/admin/lists/[e-mail]...

[William Brown]

On Wed, 2016-07-13 at 07:49 -0400, Mark Reynolds wrote:
> On 07/13/2016 02:28 AM, William Brown wrote:
> > On Wed, 2016-07-13 at 14:47 +1000, William Brown wrote:
> > > https://fedorahosted.org/nunc-stans/ticket/57
> > >
> > > https://fedorahosted.org/nunc-stans/attachment/ticket/57/0001-Ticket-57-Impl...
> > >
> > > https://fedorahosted.org/nunc-stans/attachment/ticket/57/0002-Ticket-57-Upda...
> > >
> > > May not apply without #51, #54, #56
> >
> > Mentioned it in the ticket, but this builds and works with git master of
> > DS with ldclt. So the code is "reliable" from that aspect.
>
> Can you please run the attached connection stress test script as well
> (for 24 hours)?  I'm worried that you are introducing a lot of changes
> like might cause regressions.

That's a very valid concern: my work is of course trying to improve the
reliability, but more testing is always a good idea.

I will setup an environment and run these today.

Thanks!

[William Brown]

On Wed, 2016-07-13 at 07:49 -0400, Mark Reynolds wrote:
> On 07/13/2016 02:28 AM, William Brown wrote:
> > On Wed, 2016-07-13 at 14:47 +1000, William Brown wrote:
> > > https://fedorahosted.org/nunc-stans/ticket/57
> > >
> > > https://fedorahosted.org/nunc-stans/attachment/ticket/57/0001-Ticket-57-Impl...
> > >
> > > https://fedorahosted.org/nunc-stans/attachment/ticket/57/0002-Ticket-57-Upda...
> > >
> > > May not apply without #51, #54, #56
> >
> > Mentioned it in the ticket, but this builds and works with git master of
> > DS with ldclt. So the code is "reliable" from that aspect.
>
> Can you please run the attached connection stress test script as well
> (for 24 hours)?  I'm worried that you are introducing a lot of changes
> like might cause regressions.

gstack on the process to show it really is using nunc-stans:

Thread 2 (Thread 0x7f8a17b45700 (LWP 20863)):
#0  0x00007f8a551e4863 in epoll_wait () from /lib64/libc.so.6
#1  0x00007f8a57ace803 in epoll_dispatch ()
from /lib64/libevent-2.0.so.5
#2  0x00007f8a57aba3ea in event_base_loop ()
from /lib64/libevent-2.0.so.5
#3  0x00007f8a57cfcdc5 in ns_event_fw_loop
(ns_event_fw_ctx=0x603800006380)
at /home/william/development/389ds/nunc-stans/ns_event_fw_event.c:311
#4  0x00007f8a57cf88a0 in event_loop_thread_func (arg=0x601c00158540)
at /home/william/development/389ds/nunc-stans/ns_thrpool.c:398
#5  0x00007f8a556f396b in _pt_root () from /lib64/libnspr4.so
#6  0x00007f8a584ada98 in ?? () from /lib64/libasan.so.0
#7  0x00007f8a554b6dc5 in start_thread () from /lib64/libpthread.so.0
#8  0x00007f8a551e428d in clone () from /lib64/libc.so.6

It ran for about 24 hours in total.

I ran three instances of the reliab test. No errors were encountered
during the test.

I hope that helps,