Hi,

First of all congrats for release 3.4.1 and 3.3.4!

What do you think about adding a fault injection mechanism / framework to
ZooKeeper?

I am think about something similar to:
https://issues.apache.org/jira/browse/HDFS-435

I can give it a try over the next few weeks - it's somehow related to the
research work I am doing for my master degree.

-- Andrei Savu / andreisavu.ro

I think it is a great idea. I think Pat has tried something along  
those lines a while back, and I remember it from the discussion on  
ZOOKEEPER-512. I'm not sure what happened to it, though.

-Flavio

On Dec 22, 2011, at 12:09 PM, Andrei Savu wrote:

> Hi,
>
> First of all congrats for release 3.4.1 and 3.3.4!
>
> What do you think about adding a fault injection mechanism /  
> framework to
> ZooKeeper?
>
> I am think about something similar to:
> https://issues.apache.org/jira/browse/HDFS-435
>
> I can give it a try over the next few weeks - it's somehow related  
> to the
> research work I am doing for my master degree.
>
> -- Andrei Savu / andreisavu.ro

flavio
junqueira

research scientist

[EMAIL PROTECTED]
direct +34 93-183-8828

avinguda diagonal 177, 8th floor, barcelona, 08018, es
phone (408) 349 3300    fax (408) 349 3301

I started something along these lines.

My idea was to build a non-server ZK server that could start up instantly
and would have extra methods for getting clients and injecting faults.  The
fault injection that seemed important to me was to allow the user to
"schedule" faults so that particular requests in library code could be
targeted without peeling apart the library under test.

Time constraints prevented me from producing anything useful.

On Thu, Dec 22, 2011 at 3:23 AM, Flavio Junqueira <[EMAIL PROTECTED]> wrote:

> I think it is a great idea. I think Pat has tried something along those
> lines a while back, and I remember it from the discussion on ZOOKEEPER-512.
> I'm not sure what happened to it, though.
>

Ok. I will give it a try and get back to you as soon as I have something
functional for early feedback.

-- Andrei

On Thu, Dec 22, 2011 at 5:40 PM, Ted Dunning <[EMAIL PROTECTED]> wrote:

> I started something along these lines.
>
> My idea was to build a non-server ZK server that could start up instantly
> and would have extra methods for getting clients and injecting faults.  The
> fault injection that seemed important to me was to allow the user to
> "schedule" faults so that particular requests in library code could be
> targeted without peeling apart the library under test.
>
> Time constraints prevented me from producing anything useful.
>
> On Thu, Dec 22, 2011 at 3:23 AM, Flavio Junqueira <[EMAIL PROTECTED]>
> wrote:
>
> > I think it is a great idea. I think Pat has tried something along those
> > lines a while back, and I remember it from the discussion on
> ZOOKEEPER-512.
> > I'm not sure what happened to it, though.
> >
>

There was this discussion recently: http://bit.ly/v9OQFq

A while back we had an aspectj fault injection framework for the
server, however it was not something that could be run unattended. It
basically fired failures randomly when the server called a network or
filesystem operation (every 200th call or so, it was configurable). It
verfified server (and to some extent client) error handling. I can't
seem to find it though. What Ted mentions would be more useful, esp if
you want to make it into a autonomous test(s).

Patrick

On Thu, Dec 22, 2011 at 10:44 PM, Andrei Savu <[EMAIL PROTECTED]> wrote:
> Ok. I will give it a try and get back to you as soon as I have something
> functional for early feedback.
>
> -- Andrei
>
> On Thu, Dec 22, 2011 at 5:40 PM, Ted Dunning <[EMAIL PROTECTED]> wrote:
>
>> I started something along these lines.
>>
>> My idea was to build a non-server ZK server that could start up instantly
>> and would have extra methods for getting clients and injecting faults.  The
>> fault injection that seemed important to me was to allow the user to
>> "schedule" faults so that particular requests in library code could be
>> targeted without peeling apart the library under test.
>>
>> Time constraints prevented me from producing anything useful.
>>
>> On Thu, Dec 22, 2011 at 3:23 AM, Flavio Junqueira <[EMAIL PROTECTED]>
>> wrote:
>>
>> > I think it is a great idea. I think Pat has tried something along those
>> > lines a while back, and I remember it from the discussion on
>> ZOOKEEPER-512.
>> > I'm not sure what happened to it, though.
>> >
>>