Re: Continuous buildfarm failures on hamster with bin-check

On Wed, Apr 19, 2017 at 7:03 AM, Michael Paquier
<michael(dot)paquier(at)gmail(dot)com> wrote:
> On Wed, Apr 19, 2017 at 12:48 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> writes:
>>> Tom Lane wrote:
>>>> FWIW, I'm a bit suspicious of relocating the temp stats directory as
>>>> being a reliable fix for this.
>>
>>> It's an SD card (the kind typically used in cameras and phones), not SSD.
>>> Saying it's slow is an understatement. It's *excruciatingly* slow.
>>
>> Oh, I misread it ... but still, the modern definition of "excruciatingly
>> slow" doesn't seem all that far off what 90s-era hard drives could do.
>> It is clear from googling though that there's an enormous performance
>> range in SD cards' random write performance, eg wikipedia's entry has
>> a link to
>>
>> http://goughlui.com/2014/01/16/testing-sd-card-performance-round-up/
>>
>> Seems like it's hard to judge this without knowing exactly which
>> SD card Michael has got in that thing.
>
> Professional micro SD HC 8GB, with class 10, which is utterly slow in
> the things I have tested up to now and of course I cannot find its
> read/write properties. I had a SanDisk class 10 in it that died some
> months ago, and that was way faster. RIP.

Until something happens, I have removed --enable-tap-tests from the
script of this buildfarm machine. There is no point to let it in red
without a clear resolution plan on the table first.
--
Michael