[ClusterLabs] Tuchanka

[Klaus Wenninger]

On 9/3/20 5:58 PM, Ken Gaillot wrote:
> On Wed, 2020-09-02 at 20:33 +0300, Олег Самойлов wrote:
>> Hi all.
>>
>> I have developed a test bed to test high available clusters based on
>> Pacemaker and PostgreSQL. The combination of words "test bed" was
>> given to me by a dictionary. For an russian this is rather funny, so,
>> please, tell me is this suitable phrase for this?
> As a native English speaker I had never noticed how weird that looks
> until now. :)
>
> You got me curious, but it appears the exact origin of this phrase is
> one of the few remaining things the Internet doesn't know.
>
> "Test bed" was apparently first used in the 1910s in aeronautics for
> airplane chassis used to test new equipment, so I'm going to guess it
> comes from the sense of "bed" as a flat place where things are placed,
> e.g. "sea bed" for the sea floor, "truck bed" for the back part of a
> pickup truck, "flower bed" for a small garden area.
Or like in "machine bed" more technically.
That btw. does at least in German exist as "Maschinenbett" as well.
>
>> The test bed is deployed on VirtualBox  virtual machines (VMs) in
>> MacBook Pro. Totally there will be 12 VMs which will occupy 36GiB of
>> hard disk. They will form 4 high available clusters (different
>> variants). The clusters are automatically created. And can be
>> automatically tested. The special script will in loop imitates
>> different faults, wait for restoration the cluster, fix the broken
>> node and do next test. The project is under MIT license in GitHub and
>> I just have finished translation README to English language.
>>
>> https://github.com/domclick/tuchanka
>>
>> This test bed can be used to test HA clusters. There is a list of
>> already detected problems of Pacemaker and PostgreSQL in the README.
>> And it can be used for presentations, thats why it is designed to run
>> inside one MacBook Pro. I think this will be much better instead of
>> screenshots or video to show how HA clusters survive different faults
>> in the real time.
>>
>> The software is rather outdated. It works with PostgreSQL 11 and
>> CentOS 7. The next step will be upgrading to CentOS 8 and PostgreSQL
>> 12. Please tell me, is it useful and worth to continue? Where is
>> better announce it? May be somewhere exists special mailing list for
>> such things.
> This is the right list.
>
> I haven't had a chance to look too deeply at tuchanka, but it seems
> interesting. As Jehan-Guillaume noted, there are other cluster test
> platforms already, but none of them really cover everybody's desired
> scenarios (or is easily extensible).