<div dir="ltr"><br><div class="gmail_extra"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class=""><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
One of the missing part with existing .t tests in glusterd is the client and server compatibility with different op-versions. Off late we have seen most of the bugs. I think we’d need to write tests with clients running with different versions, heterogeneous servers ( in n to n-2 range for both client and server) and ofcourse the upgrade tests.<br>
</blockquote>
<br></span>
Let me extend this thought. We need the compatibility tests run across all options that are supported and such, so that we know what breaks, a very generic compat test against an older client is not sufficient.<br>
<br>
IOW, test every option documented (and possibly not documented) in the upgrade and compat tests across servers and clients.<br>
<br>
As we hear more about large clusters, or client count explosion, we need to ensure we break nothing and can certify the same for every release.<br></blockquote><div><br></div><div>I don't disagree, but let's keep this discussion focused on specific tests and failure scenarios. Preferably, one where you can write out what needs to be tested in words and Kartik can help code/automate that.</div></div></div></div>