<div dir="ltr">In the interest of making forward progress, I went ahead and installed PostgreSQL (9.1, currently, which was what was in the repos for <a href="http://llvm.org">llvm.org</a>'s current Linux) and have set up a new 'perf' database in the LNT installation:<div>
<br></div><div>1. I'm currently doing a batch import of 2014-07's data.</div><div><br></div><div>2. I have set up the 'default' database to run a shadow import into the new database, so any new submissions should go there -- assuming the shadow import feature is still working. :)</div>
<div><br></div><div>As Tobias mentioned, we can always wipe later if it seems like a good idea.</div><div><br></div><div>Once 2014-07 has imported I will probably continue with a batch import of all of 2014's data, so that we can compare the interactive performance a bit.</div>
<div><br></div><div> - Daniel</div><div><br></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Wed, Jul 30, 2014 at 12:26 PM, Tobias Grosser <span dir="ltr"><<a href="mailto:tobias@grosser.es" target="_blank">tobias@grosser.es</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="">On 30/07/2014 20:45, Daniel Dunbar wrote:<br>
</div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="">
On Tue, Jul 29, 2014 at 9:56 PM, Tobias Grosser <<a href="mailto:tobias@grosser.es" target="_blank">tobias@grosser.es</a><br></div><div class="">
<mailto:<a href="mailto:tobias@grosser.es" target="_blank">tobias@grosser.es</a>>> wrote:<br>
<br>
On 30/07/2014 00:30, Daniel Dunbar wrote:<br>
<br>
Here are my thoughts on what we should do:<br>
<br>
<br>
Hi Daniel,<br>
<br>
thanks for your feedback.<br>
<br>
<br>
1. Before doing anything substantial, I want to get Chris'<br>
patches to<br>
rerun tests with significant changes (~= a form of adaptive<br>
sampling)<br>
landed. I have high hopes for that approach in helping making<br>
results<br>
more reliable and actionable.<br>
<br>
<br>
Adaptive sampling is a very neat idea.<br>
<br>
Out of interest. How does a database change require these changes?<br>
From my naive perspective, I would rather have a set of already<br>
running buildbots with some history to allow to understand the<br>
effectiveness of Chris' changes. Hence, having a stable database in<br>
place would be nice.<br>
<br>
<br>
It doesn't, but if the changes work well there is some value in having<br>
the data set be consistent, I thought. The other reason was that if the<br>
changes have bugs or need tweaks, it would be nicer to sort out the<br>
issues before bringing up a new database to keep things "clean" later.<br>
</div></blockquote>
<br>
We could still wipe the db after some experimental phase.<div class=""><br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
2. In the past, when bringing up new databases I have reimported<br>
some<br>
historical data using the JSON files that the server archives<br>
(as Chris<br>
noted). I could do that again here if useful.<br>
<br>
<br>
Sure. We have the last 500 builds going back to July 16 for the<br>
'clang -O3 builder'. That's 15 days history. Nothing huge, but just<br>
enough to get us history starting from the 3.5 branch.<br>
<br>
<br>
The server actually has much more data than that, I have the files to<br>
import back to 2012.<br>
</blockquote>
<br></div>
Amazing.<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="">
3. I'm not sure exactly when I will have time to bring up a<br>
PostgreSQL<br></div>
instance on <a href="http://llvm.org" target="_blank">llvm.org</a> <<a href="http://llvm.org" target="_blank">http://llvm.org</a>> <<a href="http://llvm.org" target="_blank">http://llvm.org</a>>. I<div class="">
<br>
would really love to move to a<br>
<br>
PaaS solution like Heroku to make managing this kind of thing easier<br>
(and easier to collaborate on), but we might not yet have the<br>
organizational clout for that.<br>
<br>
<br>
You seem very busy and Yi Kong has done a great job in moving LNT<br>
ahead the last months. Maybe he could help out with the installation<br>
work?<br>
<br>
<br>
Undoubtedly, the problem with having the server on <a href="http://llvm.org" target="_blank">llvm.org</a><br></div>
<<a href="http://llvm.org" target="_blank">http://llvm.org</a>> currently is we have to manage access carefully and<div class=""><br>
also be careful. If it was hosted elsewhere, we wouldn't need to worry<br>
nearly as much about changes.<br>
</div></blockquote>
<br>
Right. Maybe someone has a virtual machine to do exactly this? Or the LLVM Foundation has funds to get such a virtual machine?<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="">
4. Last I looked, the <a href="http://llvm.org" target="_blank">llvm.org</a> <<a href="http://llvm.org" target="_blank">http://llvm.org</a>><br>
<<a href="http://llvm.org" target="_blank">http://llvm.org</a>> instance mostly had<br>
<br>
polly related bots. It would be nice to start off with a more<br>
standard<br>
set of bots trying to cover the diversity of platforms, in the<br>
hopes of<br>
making the results more interesting to the larger community.<br>
<br>
<br>
Right, it is important that we get wider test coverage. On the other<br>
side, one server is just badly named:<br>
<br></div>
<a href="http://llvm.org/perf/db___default/v4/nts/recent_activity" target="_blank">http://llvm.org/perf/db___<u></u>default/v4/nts/recent_activity</a><div class=""><br>
<<a href="http://llvm.org/perf/db_default/v4/nts/recent_activity" target="_blank">http://llvm.org/perf/db_<u></u>default/v4/nts/recent_activity</a><u></u>><br>
<br>
Those are five machines building 'clang -O3' on X86 without any<br>
Polly involved. So for X86 at least one configuration is rather<br>
strongly tested.<br>
<br>
<br>
Ah, ok. Can we rename that bot?<br>
</div></blockquote>
<br>
r214324<br>
<br>
This breaks all history on the perf builder so it would be in fact a good time to introduce a new database. ;-) (Possibly with old data reimported with matching builder names).<br>
<br>
Cheers,<br>
Tobias<br>
<br>
</blockquote></div><br></div>