<br><br><div class="gmail_quote">On Mon, May 7, 2012 at 7:33 PM, Nathan Rutman <span dir="ltr"><<a href="mailto:nrutman@gmail.com" target="_blank">nrutman@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div><div><br>
On May 4, 2012, at 7:46 AM, Chris Gearing wrote:<br>
<br>
> Hi Roman,<br>
><br>
> I think we may have rat-holed here and perhaps it's worth just<br>
> re-stating what I'm trying to achieve here.<br>
><br>
> We have a need to be able to test in a more directed and targeted<br>
> manner, to be able to focus on a unit of code like lnet or an attribute<br>
> of capability like performance. However since starting work on the<br>
> Lustre test infrastructure it has become clear to me that knowledge<br>
> about the capability, functionality and purpose of individual tests is<br>
> very general and held in the heads of Lustre engineers. Because we are<br>
> talking about targeting tests we require knowledge about the capability,<br>
> functionality and purpose of the tests not the outcome of running the<br>
> tests, or to put it another way what the tests can do not what they have<br>
> done.<br>
><br>
> One key fact about cataloguing the the capabilities of the tests is that<br>
> for almost every imaginable case the capability of the test only changes<br>
> if the test itself changes and so the rate of change of the data in the<br>
> catalogue is the same and actually much less than the rate of change<br>
> test code itself. The only exception to this could be that a test<br>
> suddenly discovers a new bug which has to have a new ticket attached to<br>
> it, although this should be a very very rare if we manage our<br>
> development process properly.<br>
><br>
> This requirement leads to the conclusion that we need to catalogue all<br>
> of the tests within the current test-framework and a catalogue equates<br>
> to a database, hence we need a database of the capability, functionality<br>
> and purpose of the individual tests. With this requirement in mind it<br>
> would be easy to create a database using something like mysql that could<br>
> be used by applications like the Lustre test system, but using an<br>
> approach like that would make the database very difficult to share and<br>
> will be even harder to attach the knowledge to the Lustre tree which is<br>
> were it belongs.<br>
><br>
> So the question I want to solve is how to catalogue the capabilities of<br>
> the individual tests in a database, store that data as part of the<br>
> Lustre source and as a bonus make the data readable and even carefully<br>
> editable by people as well as machines. Now to focus on the last point I<br>
> do not think we should constrain ourselves to something that can be read<br>
> by machine using just bash, we do have access to structure languages and<br>
> should make use of that fact.<br>
><br>
</div></div>I think we all agree 100% on the above...<br>
<div><br>
> The solution to all of this seemed to be to store the catalogue about<br>
> the tests as part of the tests themselves<br>
</div>... but not necessarily that conclusion.<br>
<div> </div></blockquote><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div>
> , this provides for human and<br>
> machine accessibility, implicit version control and certainty the what<br>
> ever happens to Lustre source the data goes with it. It is also the case<br>
> that by keeping the catalogue with the subject the maintenance of the<br>
> catalogue is more likely to occur than if the two are separate.<br>
<br>
</div>I agree with all those. But there are some difficulties with this as well:<br>
1. bash isn't a great language to encapsulate this metadata<br></blockquote><div> </div><div>The thing to focus on I think is the data captured not the format. The parser for yaml encapsulated in the source or anywhere else is a small amount of effort compared to capturing the data in the first place. If we capture the data and it's machine readable then changing the format is easy.<br>
<br>There are many advantages today to keeping the source and the metadata in the same place, one being that when reviewing new or updated tests the reviewers can and will be encouraged to by the locality to ensure the metadata matches the new or revised test. If the two are not together then they have very little chance of being kept in sync. </div>
<div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
2. this further locks us in to current test implementation - there's not much possibility to start writing tests in another language if we're parsing through looking for bash-formatted metadata. Sure, multiple parsers could be written...<br>
</blockquote><div> </div><div>I don't think it is a lock in at all, the data is machine readable and moving to a new format when and should we need it will be easy. Let's focus on capturing the data so we increase our knowledge, once we have the data we can manipulate it however we want. The data and the metadata together in my opinion increases the chance of capturing and updating the data given todays methods and tools. </div>
<div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">3. difficulty changing md of groups of tests en-mass - eg. add "slow" keyword to a set of tests<br>
</blockquote><div><br></div><div>The data can read and written by machine and the libraries/application to do this would be written. Referring back to the description of the metadata we would not be making sweeping changes to test metadata because the metadata should only change when the test changes [exceptions will always apply but we should not optimize for exceptions].<br>
<br>Also I don't think 'slow' would not be part of the metadata because it is not an attribute of the test, it is an attribute of how the test is used. We need to be strict and clear here. The metadata describes the functionality of the test code and slow is not a test code function, if we want to be able to select 'slow' then we need to understand what code functionality of a test cause it to be a 'slow' test and ensure those attributes are captured. </div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
4. no inheritance of characteristics - each test must explicitly list every piece of md. This not only blows up the amount of md it also is a source for typos, etc. to cause problems.<br></blockquote><div><br></div><div>
I'm not against inheritance but the inheritance must be explicit not implicit we want to draw out knowledge about the tests if we just allow people to say 'all 200 tests in this file are X, Y, Z' then that is what will happen no one will check each test to make sure it is true and our data will be corrupted before we start.<br>
<br>So explicit inheritance might make sense, and please do propose an inheritance model for the data, we can discuss the storage format later but today let's just understand how inheritance relates to our bash tests.</div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
5. no automatic modification of characteristics. In particular, one piece of md I would like to see is "maximum allowed test time" for each test. Ideally, this could be measured and adjusted automatically based on historical and ongoing run data. But it would be dangerous to allow automatic modification to the script itself.<br>
<br></blockquote><div><br></div><div>I really do not think maximum test time as a measurement is a piece of test metadata.<br><br>Metadata describes the functionality of the test that is encapsulated within the test code itself, if the code said 'run for 60 minutes and no more' then maximum time would be an attribute. </div>
<div><br>Maybe there are a set of useful attributes like amount of storage used, or minimum clients, or minimum osts etc. etc, again these can only be metadata if they are implicitly in the test code, and for most tests they would not be definable, and the variability might be impossible to systematically capture, although I do think it's worth having a go.</div>
<div> </div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
To address those problems, I think a database-type approach is exactly right, or perhaps a YAML file with hierarchical inheritance.<br>
To some degree, this is a "evolution vs revolution" question, and I prefer to come down on the revolution-enabling design, despite the problems you list. Basically, I believe the separated MD model allows for the replacement of test-framework, and this, to my mind, is the majority driver for adding the MD at all.<br>
<div><div><br></div></div></blockquote><div>Database is good and I believe metadata in the source fulfils that objective whilst being something that we can manage with what we have today manually, whilst easily creating tools for some automation. When we do begin work on a new test framework approach we will have all the data at hand to be manipulated in any way that we want, including if we want separating it and storing it somewhere else.<br>
<br>I don't think creating the metadata however is linked with a new test-framework, creating the metadata is required because today we do not know what we have and we need to know what we have today whatever strategy we use for the future.</div>
<div><br></div><div>Chris</div></div><br>